the impact of tax reform in modern dynamic …judd/papers/aeifin.pdfthe impact of tax reform in...

The Impact of Tax Reform in Modern DynamicEconomies

Kenneth L. Judd∗

Hoover Institution, Stanford, CA [email protected] Bureau of Economic Research

A theoretical case for consumption taxation, based on the inverse elasticity andproductive efficiency principles from optimal tax theory, is presented foreconomies with perfect competition. Tax reform analyses often ignore issues ofimperfect competition, risk, and human capital. These are all important elementsof any modern economy. Our theoretical arguments are then extended to allow forimperfect competition, risk, and human capital. Including these features in ouranalysis at least doubles and often triples the estimated gains from switching toconsumption taxation. Furthermore, the gains from consumption tax reform aremore evenly distributed than conventionally thought.

∗I thank Alan Auerbach, Kevin Hassett, Glenn Hubbard, Alvin Rabushka, and partic-ipants of the AEI�s �The Transition Costs of Fundamental Tax Reform� for their manyuseful comments. I acknowledge the support of NSF grant SBR-9708991.

The Impact of Tax Reform in Modern Dynamic Economies 1

Tax policy in the United States since WWII has been based on theprinciples of an income tax. Its intellectual foundation lies in the Haig-Simons approach to income taxation - deÞne income properly and tax it.However, economists over the past thirty years have increasingly argued formoving away from income taxation and towards consumption taxation. Taxreform debates often focus on the choice between income and consumptiontaxation.

The key issue is the taxation of savings and investment1. Many the-oretical analyses have argued for a zero long�run capital income tax rate.Early arguments, such as Feldstein (1978), Atkinson and Sandmo (1980),Auerbach (1979), and Diamond (1973), relied heavily on separability as-sumptions and identical agents in each cohort. Judd (1985b) proved thatthe optimal long-run capital income tax rate is zero even when tastes arenot separable and agents have different tastes and abilities. Others have ex-plored taxation issues in models of economic growth. Eaton (1981) showedthat capital income taxation reduces an economy�s long-run growth rate andHamilton (1987) demonstrated that asymmetric treatment of different kindsof investment has a high efficiency cost. Judd (1999) generalized the analysisin Judd (1985b) to include human capital investment, government expen-diture, and various forms of growth. All of these analyses argue stronglyagainst taxation of asset income in the long run.

The increasingly robust theoretical case against asset income taxationhas been supplemented by estimates of how much the economy would bene-Þt from tax reform. Studies such as Jorgenson and Yun (1990) and Auerbach(1996) show that switching to consumption taxation would signiÞcantly in-crease savings and labor supply, and improve productivity. Jones et al.(1993) uses computed examples to show that asset income should be smalleven in the intermediate run. Both theoretical and empirical work showsthat a pure income tax system is far from best in terms of aggregate output.

The U.S. tax system has also evolved into a hybrid system combiningfeatures of income and consumption taxation2, but the presence of the cor-porate income tax and the limited nature of savings incentives still givesthe current tax system a strong income tax ßavor. Most economists agreethat moving completely to consumption taxation would improve aggregate

1We must immediately clear up a semantic problem which can arise in discussing thetaxation and nontaxation of asset income. In this paper, any comment on whether a taxsystem taxes asset income implicitly refers to the effective tax rate on new investment.In this sense, the current tax system taxes asset income, but the Hall-Rabushka Flat Taxand most other consumption tax proposals do not tax asset income.

2See Aaron et al. (1988) for a descriptions of the problems of a hybrid tax system.


productivity and income in the long-run. Problems arise when we look attransition and distributional issues. Some critics have argued that equityconsiderations and transition problems related to changes in asset pricesblunt the case for a complete move to consumption taxation and make itpolitically less viable. In particular, the elimination of many middle-classtax deductions reduce middle-class support for tax reform. Possible adverseimpacts on asset prices may make some individuals, particularly the elderly,worse off than under the current tax system. Any debate on tax reformwill consider the trade-offs between the long-run beneÞts and the short-runtransition problems.

This study examines the conceptual basis for a consumption tax andintroduces many features, previously ignored in tax reform analyses, of amodern economy which substantially strengthen the case for switching com-pletely to consumption taxation. Despite the theoretical literature, someauthors (e.g., Gravelle, 1994) still assert that efficient taxation of capital de-pends on the relative elasticities of consumption demand and labor supply.We review the theoretical ideas behind the consumption tax and show thatthe case against capital taxation and for consumption taxation is surpris-ingly robust, not depending on unknowable, technical details of the economy.This conceptual foundation then leads us to explore other aspects of con-sumption versus capital taxation. In particular, we explore the implicationsof adding imperfect competition, risky assets, and human capital formationto the standard analysis. Any analysis, including that in this paper, mustmake many simpliÞcations, and ignoring those elements was natural for ini-tial analyses of tax reforms. Now that we understand the implications of taxreform in a competitive economy, we should extend our models and makethem more realistic. It is natural to include imperfect competition, riskyassets, and human capital in tax analysis since it is difficult to imagine amodern dynamic economy without these features.

Most will not be surprised that adding imperfect competition, risky as-sets, and human capital affects our results, but this study argues that incor-porating these elements substantially strengthens the case for a consumptiontax. First, including these elements of a modern economy substantially in-creases our estimates of the gains to long-run productivity. Interactionsbetween taxation and imperfect competition increase the welfare cost of in-come taxation. The current U.S. tax system discriminates against risky as-sets; we show that any tax reform which eliminates this feature will producesigniÞcant efficiency gains. Including human capital in our analysis increasesthe welfare gains of eliminating taxation of income on new investment.

Second, these extra considerations also reduce transition problems. The


incorporation of imperfect competition reduces, possibly even reverses, ad-verse movements in asset prices. This change, plus a detailed view of U.S.demographics, reduce the problems of protecting older individuals who maynot live long enough to enjoy the long-run beneÞts of tax reform. The in-corporation of human capital also suggests new ways, consistent with theconsumption tax principle, to compensate the middle class for the elimina-tion of current deductions.

We motivate these considerations by reviewing some basic ideas frompublic economics and industrial organization. In particular, we present theinverse elasticity and production efficiency results from optimal tax theory,use them to analyze the inefficiencies of conventional income tax, and discussinteractions between taxation and imperfect competition. Conventional dis-cussions focus on the distinction between income and consumption taxation.We argue that there is really no distinction between income and consump-tion taxation since income taxation is really a special pattern of consumptiontaxation. More precisely, we argue that income taxation is a particularlybad form of consumption taxation violating basic rules for a good tax sys-tem. This paper argues that the focus should instead be on the taxationof consumption today versus consumption tomorrow, and on the taxationof intermediate goods versus the taxation of Þnal consumption goods. Thisfocus helps explain old results and point in useful new directions.

First, there are many tax-like distortions in the private sector. Wheneconomics professors teach competitive economic theory they often use theexample of the hundreds of thousands of farmers producing some agriculturalproduct, and correctly argue that no individual producer has any impactover the price of his crop. This competitive paradigm is the one usually em-ployed in tax reform analysis. While the competitive model may have beena valid simpliÞcation in 1798, it is certainly not in the modern industrialhigh-technology U.S. economy of 1998. Today imperfect competition andoligopolistic interactions provide a more appropriate description of much ofthe economy, and is particularly appropriate when discussing capital goodsand innovations which are sources of economic growth. Some of the ideas ofcompetitive theories still hold. In particular, competitive forces in oligopolis-tic sectors may reduce proÞts to competitive returns and prices to averagecost. However, we expect prices to exceed marginal cost. Efficiency and wel-fare is determined by the relation between price and marginal cost, not priceand average cost. This wedge between price and marginal cost is essentiallya tax, even when it is generated by the private economy.

We show that the presence of imperfect competition strengthens thecase for consumption taxation since it increases our estimates of the aggre-


gate efficiency gains of tax reform. In fact, we show that estimates of thediscounted welfare gains from switching to a consumption tax are at leastdoubled for central estimates of the critical parameters, and the estimatesof the long-run gains are even greater.

Second, risk is usually ignored in tax analysis. This is potentially animportant problem since the current income tax discriminates against riskyequity investment in favor of safe debt investments. This discriminationappears to violate optimal taxation principles: if both risky and safe as-sets produce income for future consumption, why should the tax systemdiscriminate between alternative investment strategies? Consumption tax-ation would eliminate this discrimination, improving both the allocation ofcapital and incentives to save. Even some partial reforms would be of sub-stantial value. We show that eliminating the debt-equity distinction in thetax code may by itself achieve half the beneÞts of moving completely to aconsumption tax.

Third, tax analyses usually focus on labor supply and physical capitalformation. Since human capital is more important than physical capital ina modern economy, this is a serious limitation. Many economists argue thatthe current tax and education system puts little tax burden on human cap-ital formation, a position which would seem to justify the focus on physicalcapital taxation. We make two points. First, we show that adding humancapital formation to our analysis increases the estimated beneÞts from taxreform even if human capital investment incentives are undistorted. Second,we argue against the conventional view, pointing to the large amount of ed-ucational expenditures, both private and public, which would be includedin the tax base by most tax reform proposals. This fact violates the con-sumption tax principle since a true consumption tax would deÞne the taxbase to be output minus all investment expenditures.

These three considerations, imperfect competition, risk, and human cap-ital accumulation, all indicate that consumption taxation is even more ben-eÞcial, both in the long run and along the transition, than conventionallyargued. These points initially ignore distributional problems. Since dis-tributional concerns are important to any political advocacy, we make twoimportant points. First, some distributional analyses argue that the elderlymay lose from tax reform. A switch to consumption taxation may causethem to pay new taxes, either directly or implicitly through a decline inasset values, on their wealth. In particular, Gravelle (1995) predicts a 20-30% fall in stock prices if the Hall-Rabushka Flat Tax is passed. Thesearguments typically assume perfect competition where no Þrms earn anyeconomic rents. While farms and other small businesses may be competi-


tive, they are not part of anybody�s stock portfolio. It is difficult to viewÞrms like Microsoft, GM, and Boeing as perfectly competitive price takers.I also suspect that their CEO�s would not last long in their jobs if theywere satisÞed with normal proÞts and did not pursue opportunities to earnextranormal proÞts for their shareholders. We argue that any predictionsof asset price collapses are blunted, possibly even reversed, when we includeimperfect competition in the analysis. The presence of imperfect competi-tion implies that Þrms earn pure proÞts on extra production, and that theincrease in future output induced by the Flat Tax (or any other consumptiontax) would cause asset prices to immediately rise. This asset price increasewould allow elderly asset holders to participate now in the future beneÞtsof tax reform, and make tax reform more uniformly beneÞcial across thegenerations.

Second, many middle class families would lose from tax reform. Two rea-sons for this are the loss of deductions for home mortgage interest and stateand local taxes. Some propose keeping the mortgage interest deduction toavoid middle class losses and get them to join the political coalition for a con-sumption tax. This would substantially reduce the potential efficiency gainsof tax reform since it would continue the current bias against nonresidentialbusiness investment. An alternative adjustment in tax reform proposals isto allow the deductibility of some educational expenditures. This wouldhelp with the distributional issues since such deductions could be aimed atmiddle class taxpayers, but would not deviate from the consumption taxprinciple.

Many consumption proposals have been put forward, including those de-scribed in Bradford (1986), Hall and Rabushka (1995), McLure and Zodrow(1996), and Weidenbaum (1997). These ideas also apply to any VAT ornational sales tax proposal, since they would eliminate taxation of incomeon new investment. I do not focus on any one proposal since the argumentsfor consumption taxation made here apply to all of them. Other propos-als argue for eliminating the double taxation of equity income through theintegration of individual and corporate taxation, thereby eliminating theasymmetric treatment of equity and debt assets; see U.S. Treasury (1992).Many of our results also apply to those proposals since we will be focusingon capital income taxation. Similarly, our arguments apply to more conven-tional tinkering of the tax code such as the reintroduction of the investmenttax credit. Our results show how important it is to include imperfect com-petition, risk, and human capital formation to the analysis of any tax reformproposal.

The case for consumption taxation is strong, and made stronger when


we include those features which make our economy a modern and technolog-ically advanced one. Recognition of these elements should help us achievesubstantive tax reform.

1 Evaluating Alternative Tax Systems

This section presents the conceptual foundation for our arguments. Any taxsystem will produce distortions and damage economic performance. Thetask of policymakers is to choose a tax policy which produces the least dam-aging pattern of distortions. This task is particularly difficult in a dynamiceconomy where one needs to trade off distortions today against their futureconsequences.

The arguments in this paper rely on two basic results from optimal taxtheory plus an argument from monopolistic competition theory. First, theinverse elasticity rule3 argues that the tax on a good should be inverselyproportional to its demand and supply elasticities. We show how to applythat rule to dynamic contexts, and why an income tax is really a particularlyinefficient kind of consumption tax.

Second, the productive efficiency principle of Diamond and Mirrlees ar-gues against the taxation of intermediate goods, such as capital. The currenttax system discriminates in favor of capital in the form of owner-occupiedhousing and against capital used to produce other goods. It also treats hu-man capital and physical capital differently even though both are essentiallyintermediate goods. Financial structure is also a type of intermediate goodsince debt and equity have no direct consumption value, but the current U.S.tax system discriminates against equity and in favor of debt. The productiveefficiency principle helps us understand what a true consumption tax wouldlook like and why deviations from the productive efficiency principle are sodamaging to economic efficiency.

Third, we display similarities between taxation and imperfect competi-tion. Any Þrm which has some control over the price it charges for its goodswill charge a price in excess of marginal cost. That gap is similar to a tax.Recognizing the presence of imperfect competition is similar to recognizingthe presence of other taxing authorities. The presence of these other �taxes�will signiÞcantly affect our view of the government�s taxes.

3See Atkinson and Stiglitz (1972) and Atkinson and Sandmo (1980) for formal presen-tations of optimal taxation theory.


Figure 1: Inverse Elasticity Rule

1.1 The Inverse Elasticity Rule and Asset Income Taxation

The inverse elasticity rule says that the optimal tax on a commodity isinversely proportional to its demand elasticity4. We can illustrate this byconsidering the demand curves for two goods displayed in Figures 1. Bothgoods are assumed to have a constant unit marginal cost. The demand curvefor good 1 in the left half of Figure 1 displays the impact of a tax equal toτ1. The box R1 is the revenue raised by the tax, CS1 the consumer surplus,and H1 the efficiency cost of the tax. Demand for good 2, displayed in theright half of Figure 1, is assumed to have be less elastic. If we imposed thesame tax rate of τ1 on good 2 the revenue is R2 and the welfare cost is H2.Since the demand for good 2 is less elastic, the optimal policy is to tax good2 at a higher rate, say τ2. This higher tax increases revenue by an amountequal to the area in box A minus the area in box C. The extra efficiency costis B + C. The objective is to equate the marginal cost of a higher tax perdollar of revenue across different goods. This is accomplished by imposinghigher taxes on the less elastically demanded goods. In Figure 1 we wouldset the tax on good 1 at τ1 and choose a higher tax of τ2 on good 2.

The inverse elasticity rule may seem to have little application to dis-cussions of income taxation and savings. However, it is the best way toview income taxation. Suppose that the different goods in Figure 1 repre-sent consumption of goods and leisure at different dates. Income taxationimplies a pattern of distortions across consumption and leisure at variousdates. For example, if we save some money at time 0 for consumption attime t, then a tax on investment income essentially taxes consumption attime t. Suppose r is the before-tax interest rate, and τ is the interest tax

4We ignore supply elasticities in this discussion since they are not as relevant for ourapplications of the inverse elasticity idea.


rate. The social cost of one unit of consumption at time t in units of the time0 good is (1+ r)−t and the after-tax price is (1+ (1− τ)r)−t. This impliesa tax distortion between MRS, the marginal rate of substitution betweentime t consumption and time 0 consumption, and MRT , the correspondingmarginal rate of transformation, equal to

MRS

MRT=

µ1+ r

1+ (1− τ)r¶t

(1)

This distortion is the same as if we taxed consumption at time t at the rate

τ∗c =µ

1+ r

1+ (1− τ)r¶t− 1 (2)

The key fact illustrated by this formula is that the commodity tax equivalentis exploding exponentially in time!

The situation is displayed in Figure 2. Figure 2 shows the demandfor the time t consumption good relative to some untaxed good c0 (suchas time 0 leisure). This income tax is equivalent to a commodity tax ontime t consumption equal to τ∗c per unit of the time t good. We make thecommon assumption that the consumption demand curves are identical andindependent across time and not affected by leisure. The optimal tax systemwould impose the same tax on consumption at each different time. Instead,a constant positive interest tax is equivalent to an exponentially growing taxon time t consumption, strongly violating the inverse elasticity rule. Noticein Figure 2 that as t increases, the deadweight loss triangle, H, grows andsqueezes the revenue box, R.

The exponential explosion in (2) appears dramatic, but we need to checkthat it is quantitatively important over reasonable horizon. Table 1 displaysthe consumption tax equivalents, τ∗c , for various combinations of r and τ . Wesee that the results depend substantially on the magnitude of r. For r = .01,the mean real return on safe assets, the effects are small. For example, evena 50% tax on interest income implies only a 22% tax on consumption tax 40years hence compared to a .1% tax on consumption a year away. However,the situation is much different when r = .10. When τ = .3 (which is lessthan the tax rate on equity-Þnanced capital), the effective consumption taxover a one-year horizon is 3%, but it is 59% over a ten-year horizon, and awhopping 543% over a 40-year horizon! It is hard to imagine any governmentpassing a 59% sales tax in 2008, but that is effectively what we do to manyinvestors if we continue with an income tax system between 1998 and 2008.


Figure 2: Commodity Tax Equivalents of Asset Income Taxation

Table 1: Consumption Tax Equivalents, τ∗cr τ t

1 5 10 20 30 40

.01 10 .1 .5 1 2 3 430 .3 1 3 6 9 1350 .5 2 5 20 16 22

.10 10 1 5 10 20 31 4430 3 15 32 74 129 20250 5 26 59 154 304 543

The implications of this analysis are clear. If utility is separable acrosstime and between consumption and leisure, and the elasticity of demand forconsumption does not change over time, the best tax system would have aconstant commodity tax equivalent. This can be accomplished by a con-stant consumption tax. However, any nonzero asset income tax producessubstantial violations.

While the exposition above focuses on special cases, the result is robust.The results in Judd (1985b, 1999) show that the optimal tax on asset incomeis zero in the long-run, even when preferences are far more general than thoseused in dynamic tax analyses. The key idea is that exploding consumptiontax rates are not efficient and that the explosion is quantitatively important.

This result is not just an aggregate result assuming everyone is the same.It is true for each individual if his tastes do not change signiÞcantly overtime. Therefore, even if tastes vary across individuals, each individual will


prefer a constant consumption tax over an income tax which extracts thesame revenue from him.

The inverse elasticity rule argues for a different tax on all goods, whereasconsumption tax proposals actually propose a single tax rate. While thismay appear to be a serious difficulty, we will ignore it. This approach issupported by the arguments in Balcer et al.(1983). They show that while anoptimal commodity tax system would have very different rates across goods,there is little welfare difference between that tax system and a revenue-equivalent ßat tax. Given the extra complexity and administrative cost ofa tax system which charges different tax rates on different goods, it seemssensible to stay with a uniform consumption tax.

This analysis does not necessarily imply that there should be no taxationof asset income. Suppose that tastes depend on age. If we assume that theelasticity of demand for consumption fell with age in just the right way,then a constant interest rate tax would be optimal; this would require thedemand curve in Figure 2 for the time t good to become less elastic as tincreases. Such an age-dependence could result if one had just the rightinteraction between consumption demand and leisure. However, I have notseen advocates of asset income taxation use this approach5. My suspicionis that these arguments would be very fragile since our knowledge of thecritical elasticities is too imprecise for such a purpose. In any case, it ishard to imagine demand elasticities changing enough to justify substantialasset income taxation. In particular, Table 1 tells us that to justify a 30%income tax if r = .1 over a twenty year horizon, we would need consumptionelasticity to fall by a factor of 25 over those 20 years, a rather implausiblesituation. Therefore, the constant elasticity case is a reasonable one to use.

The distinction between factor income taxation and commodity taxationis misleading since none of the problems in Figure 2 applies to wage incometaxation. If τL is a constant wage tax and τK a constant interest rate tax,the MRS/MRT distortion between time 0 consumption and time t leisureis µ

MRS

MRT

¶c0,`t

=

µ1

1− τL¶ µ

1+ r

1+ (1− τK)r¶t

(3)

Equation (3) represents how taxes distort decisions to sacriÞce consumptionat time 0 to gain extra leisure at time t. This distortion also grows overtime but only because of the interest rate tax. Wage taxation does not

5Simulations of tax policy analysis may stumble on this if they assume tastes whichlead to time-varying consumption demand elasticities.


aggravate the distortions in savings but asset income taxation does aggravatedistortions between consumption and leisure at different dates.

Our commodity tax interpretation of factor income taxation and theinverse elasticity rule reveal many features of factor taxation. This approachshows us how distortionary asset income taxation is, and hints at the valueof removing it from tax systems.

1.2 The Productive Efficiency Principle

The second important principle is the Diamond-Mirrlees result about pro-ductive efficiency. The essential argument is that a tax system may unavoid-ably cause distortions in consumption, but there is no need to also force theeconomy to produce that output in an inefficient fashion. The chief implica-tion of the Diamond-Mirrlees efficiency result is that an optimal tax systemwould tax only Þnal goods, not intermediate goods.

For example, we may want to tax clothing and meat, but we do notwant to tax sewing machines and meat storage lockers. If we taxed sewingmachines, clothing producers would substitute away from mechanical pro-duction and towards labor-intensive methods, reducing the productivity ofthe economy. Even if we wanted to tax clothing more heavily than meat,any differential treatment of sewing machines andmeat storage lockers wouldjust distort the allocation of capital. Taxes on sewing machines and meatlockers are all ultimately paid by consumers in any case. It is better torely on direct taxation of clothing and meat consumption and allow theproduction of both to proceed undistorted by taxation of capital inputs.

The productive efficiency principle applies to any analysis of income tax-ation since capital goods are intermediate goods. In fact, taxation of capitalgoods is equivalent to sales taxation of intermediate goods. This can beseen by noting, for example, that a 100% sales tax on capital equipment isequivalent to a 50% tax on the income ßow from that capital equipment.Since intermediate good taxation will generally reduce the productivity ofan economy, capital income taxation will likely produce similar factor dis-tortions, particularly if there are many capital goods.

When we combine the productive efficiency principle with the inverseelasticity principle, we arrive at a strong case against capital income taxa-tion. Differential taxation of capital goods will produce inefficiencies in theallocation of productive inputs. A uniform tax on capital inputs may notdistort allocation but will effectively create an exploding consumption taxas illustrated in Table 1. Therefore, an optimal tax structure would tax onlyÞnal goods.


The productive efficiency principle is well-recognized in tax reform argu-ments. One of the key beneÞts from consumption taxation is the eliminationof differential taxation across various capital goods; see Auerbach (1989) andGoulder and Thalmann (1993) for recent examinations of the importance ofproductive efficiency. The changes in 1986 attempted to create uniformtaxation across capital goods. Auerbach (1989) argues that any optimaldeviations are small under perfect competition.

The Diamond-Mirrlees principle does rely on special assumptions, lead-ing some to argue against its relevance in tax discussions. Two provisosimmediately come to mind. First, Diamond and Mirrlees assume each com-modity is taxed at a separate rate. Again, as we indicated above, we do notview that as a serious problem. While Balcer et al. (1983) did not considera general equilibrium case where intermediate goods could be taxed, we sus-pect that their conclusion that uniform taxation is almost as good as theoptimal nonuniform tax is robust. Second, the productive efficiency resultalso assumes that all pure proÞts are taxed away, whereas pure proÞts arenot taxed away in the current tax system nor in any proposed reform. Infact, the drop in marginal rates from most reforms would reduce the taxa-tion of pure rents. We will show that this is also not a serious impedimentto applying the production efficiency principle when we consider plausibleestimates for tastes and technology.

This Diamond-Mirrlees productive efficiency principle provides us with atheoretical basis for consumption taxation. However, it also tells us that weneed to pay careful attention to what is an intermediate good and what is aÞnal good. This distinction will play a critical role below in our discussionof human capital.

1.3 Imperfect Competition and Taxation

The third idea we use is that taxation decisions of the government and thedistortions produced by imperfect competition in the private sector are sim-ilar in their implications. A Þrm which charges a price above marginal costis effectively acting as a tax collector. Any national consumption or incometax is imposed on top of any private sector distortions. This accumulationof distortions will substantially affect our estimates of the burden of taxesand our relative evaluation of consumption and income tax systems.

The key principles are displayed in Figure 3. Suppose that a good is notsold at its marginal cost, equal to 1 in Figure 3, but is sold at a markedup price, 1+m. This markup can arise and be sustained for many reasons.The producer may have market power because of large Þxed costs of entry


or because his product is differentiated from the products of competitors.Alternatively, the producer may hold a patent which makes him a legalmonopolist.

Any markup above marginal cost acts essentially as a tax. In Figure 3,Hm, is the efficiency cost of such a markup, just as H1 and H2 were theefficiency costs of taxation in Figure 1. The box P +Hτm in Figure 3 is themonopolists �tax revenue� constituting proÞts in excess of economic costs.The economic effects of any markup is similar to taxation since both causethe buyer to pay a price in excess of the true marginal cost. These two casesdiffer in who receives the markup, the government in the case of a tax anda private Þrm in the case of a markup. Both taxation and markups createsefficiency losses and rents.

We will rely heavily on this analogy between taxation by the govern-ment and markups arising from imperfect competition. This analogy isparticularly appropriate in the case of patents. The holder of a patent is notnecessarily a monopoly producer. In fact, many patent holders do not pro-duce their product. The key feature of a patent is that the patent holder canimpose a tax on the purchase of the patented good, either directly throughproducing the good and charging a price in excess of marginal cost, or indi-rectly through a royalty. These distortions reduce economic efficiency andlead to underproduction of the patented good, but are justiÞed by the incen-tives they create for innovation. Without the rents produced by a patent,an innovator may not have sufficient incentive to undertake the Þxed costsof research and development, a situation leading to an even worse situationof no production of a desirable product. Therefore, even though patent mo-nopolies reduce efficiency just like taxes do, we do not want to destroy therents they create.

The patent monopoly story is the simplest one we can use to illustratethe key arguments, but our points are more robust and apply to any contextwhere Þrms charge a price in excess of marginal cost. In many cases thesemarkups occur due to product differentiation and increasing returns to scale,situations which share many features of a patent monopoly even if there isno formal property right. Our analysis revolves around the presence of amarkup of price over marginal cost,.whether it arises from patent monopoly,an oligopoly of differentiated competitors, or some other form of imperfectcompetition.

Markups may also occur due to collusion or corruption, but that is theconcern and responsibility of antitrust policy. Our arguments apply to im-perfect competition which remains after appropriate application of antitrustlaws. We do not argue that tax policy is a substitute for antitrust policy. In-


Figure 3: Taxation and Monopolistic Competition

stead, we argue that tax policy should take notice of the fact that imperfectcompetition is an important part of any modern economy.

Suppose that we introduce a tax τ into an imperfectly competitive mar-ket. The buyer now pays both the markup and the tax, resulting in a totalprice of 1+m+ τ . The m+ τ portion acts as a tax, raising price above themarginal cost and producing revenues now for the government. In this case,the government�s revenue is the box R and the Þrm�s proÞts are P . Thetax τ causes the monopolist to lose the box Hτm in proÞts and causes theconsumers to lose Hτ in consumer surplus. The cost of the tax is not justa triangle of consumer surplus but also a box of pure proÞts. The efficiencycost of the tax is now larger relative to the revenue raised because of thepre-existing distortion.

Joan Robinson (1934) noticed these facts and argued that a good taxpolicy would use subsidies to bring buyer price down to social marginal cost.This would imply that in Figure 3 we would want to pay the buyer a subsidyequal to the markup m. She also argued that this policy would have someundesirable effects since it would increase monopoly proÞts and likely beregressive in its impact on income distribution. Since it would be difficultto tax away these extra proÞts, she did not endorse such an approach.


We argue that these distributional concerns are not important in theU.S. economy. In modern dynamic economies, it is difficult for a Þrm tomaintain large monopoly rents. High proÞts encourage entry by imitators.We used to think of IBM as a Þrm with large market power before it washit by competition from competing producers of personal computers andworkstations. For many Þrms, the current proÞts arising from setting pricesabove marginal costs are necessary to recover R&D costs and other Þxedcosts of production. This view of monopolistic competition is supported byHall(1986) who Þnds that there is little evidence of supernormal returns toÞrms even though he Þnds that prices substantially exceed marginal costs.

Before continuing, we should note the limited way in which we will usethese imperfect competition ideas. The key idea here is that pre-existingdistortions increase the efficiency cost of governmental taxation, even if taxpolicy is not used to Þne-tune those distortions. We will see below how thislimited argument strengthens the case for consumption taxation.

1.4 Taxation in a Simple Dynamic Competitive Model

We will use some standard analysis to illustrate the signiÞcant beneÞts ofmoving away from asset income taxation and towards consumption taxation.We assume the simple growth model used in Judd (1987). The key featuresin Judd(1987) are that output is produced by capital and labor, and isdivided into consumption and investment. There are no adjustment costs,and we use the representative agent paradigm. We assume a Cobb-Douglasproduction function with capital share .25. We assume that labor supply hasa compensated elasticity equal to η > 0 and that the consumption demandelasticity6 is γ > 0.We assume a proportional tax on labor income at a rateof τL and a proportional tax on capital income at rate τK .

Table 2 displays the marginal efficiency cost of various tax changes forvarious values of γ and η. We assume that the economy begins with onetax policy and makes small changes in labor or capital income taxation, orintroduces a small investment tax credit (ITC) applied to all investment.We do not explicitly include a consumption tax, but an increase in an ITChas the same effect of reducing the effective tax on new capital withoutreducing the taxation of old capital. For example, the Flat Tax proposesexpensing of capital expenditures, a measure which is equivalent to a largeITC. These three policy tools cover most of the policy options used in thepast and proposed for the future.

6See Judd, 1987, for a long list of empirically estimated labor supply elasticities andlabor tax rates used there to compute MEB.


We Þrst examine the case where τL = τK = .3 initially, and then weexamine the case where the economy begins with τL = .4 and τK = .5.MEBL is the marginal loss of utility (measured in dollars) per dollar ofrevenue raised if τL is increased. MEBK (MEBITC) is the correspondingindex for increases in τK (an ITC). The MEB indices in Table 2 are dis-counted present values which include the transition process from one taxpolicy regime to another. We expect the MEB > 0 since we expect thatany tax policy change which raises revenues will reduce utility; however,MEB < 0 is possible in severely distorted systems.

Table 2: Efficiency Costs of Various Policy Changes

τL = .3, τK = .3 τL = .4, τK = .5γ η.1 .1.1 .4.1 1.0.5 .1.5 .4.5 1.02.0 .12.0 .42.0 1.0

MEBL MEBK MEBITC.02 .15 .46.04 .21 .72.05 .25 .99.04 .36 1.9.11 .42 2.6.19 .48 4.0.05 .69 −240..20 .91 −6.9.50 1.3 −3.5

MEBL MEBK MEBITC.04 .38 1.1.07 .51 1.8.08 .62 2.5.07 1.3 −15..21 1.5 −9.8.35 1.7 −7.3.09 5.8 −2.8.39 22. −2.41.19 −11. −2.0

Table 2 illustrates several important points. First, we do not have agood quantitative grasp of the welfare costs of tax changes. The values ofcritical parameters used in Table 2 are all in the range of existing empiricalestimates. It is difficult to choose among the empirical estimates of γ and ηsince they differ in terms of data sets and estimation strategy. The typicalcalibration approach would argue vigorously for one particular parameterchoice and ignore others. I am sceptical about our ability to make suchchoices given the noisy data we have and the enormous gap between thissimple model and the far more complex real world.

Second, Table 2 shows us that we do not need good estimates to rankalternative tax policy changes. In all cases in Table 2, replacing capital in-come taxation with labor income taxation would improve welfare, usuallyby a substantial amount relative to the revenue shift. Furthermore, changeswhich focus on encouraging new investment, such as the ITC, are particu-larly effective in improving economic performance with small revenue loss.In fact MEBITC is sometimes negative, implying that an increase in the


ITC would raise revenues because the extra capital income tax revenue onthe new capital and the extra wage taxation from the higher wages wouldpay for the costs of the investment tax credit. An increase in the ITC issimilar to the introduction of a Flat Tax. Both reduce taxation of new in-vestment but do not reduce the tax burden on old capital. In fact, the FlatTax can be viewed as an income tax at rate τ and plus an ITC at rate τand no depreciation allowances.

Third, the more elastic the labor supply, the greater the difference be-tween MEBL and MEBK . A static perspective suggests that the relativecosts of labor and capital income taxation depends on the elasticities of sav-ings and labor supply, and that as the labor supply elasticity increased, thewelfare cost of labor taxation relative to capital income taxation would rise.The opposite is true in Table 2 where both MEBK and MEBITC rise evenmore rapidly thanMEBL as we increase labor supply elasticity, η. The res-olution of this puzzle lies in our MRS/MRT distortion expressions abovein (1) and (3). These expressions tell us that asset income taxation impliesan exploding distortion for both consumption and leisure demand. As thelabor supply elasticity rises, the importance of this labor market distortionalso rises, and the total distortion due to asset income taxation rises.

The τL = τK = .3 case for the initial tax policy is low relative to thecurrent U.S. tax system. The case of τL = .4 and τK = .5 is closer to theconventional description of the tax system before 1981, but is not generallyconsidered descriptive of the current tax system. Of course, the welfarebeneÞts of tax reduction is much greater when we begin with higher taxrates. We will also see that the τL = .4 and τK = .5 scenario is actuallyplausible when we consider the impact of imperfect competition.

The robustness of the results in Table 2 is surprising since we normallyexpect the results from computational general equilibrium to depend criti-cally on the elasticity parameters. The magnitudes of the MEB indices dodepend on elasticity values, but the ranking of alternative policies does not.This indicates that something fundamental is behind the results. We arguethat the critical facts come from optimal tax theory: taxation of asset in-come corresponds to exploding commodity taxation but labor taxation andconsumption taxation do not.

1.5 Optimal Tax Theory and Tax Reform

Before continuing, we will summarize our theoretical arguments. It is impor-tant to do this since our results strongly contradict the standard intuitionused by many in the tax reform literature.


Gravelle�s (1994) discussion of the welfare effects of consumption versuscapital income taxation is a good statement of the commonsense approach.She asserts7

Theory does not tell us, a priori, whether eliminating capital in-come taxes will increase overall efficiency, since it reduces onedistortion at the price of increasing another. ... The efficiencyeffects depend on assumptions about behavioral effects. If indi-viduals are relatively unwilling to substitute consumption overtime and relatively willing to substitute leisure for consumptionof goods, then a signiÞcant tax on capital income would consti-tute part of an optimal tax system. These behavioral effects aredifficult to estimate empirically.

This intuition is a natural one. Its references to substitution propensitiesappear to invoke the inverse elasticity rule we also invoke, and argues that wemust accept trade-offs among various distortions. However, the argumentswe have made do not make any qualiÞcations concerning the relative elastic-ity of intertemporal substitution and labor supply. Separability assumptionswere made in some earlier analyses, but even those assumptions are absentin Judd (1985b). Many of the analyses arguing for no long-run taxationof capital assume a constant intertemporal elasticity of consumption, butthat focus is not restrictive. Table 1 shows that even a small capital incometax implies rapidly exploding consumption tax equivalents, and there is noevidence that individual consumption elasticities vary enough to make sucha tax policy efficient. Plausible values for consumption demand and laborsupply elasticities offer no support for asset income taxation in the long run.

This discussion ignores the transition process, but there again we Þnd noevidence supporting asset income taxation on efficiency grounds. Table 2 infact shows the opposite. The gapsMEBITC−MEBL andMEBK−MEBLrepresent the efficiency gain from increasing labor taxation and using therevenues to Þnance an increase in the ITC or a decrease in capital incometaxation. Table 2 shows that this gain increases as we increase our estimateof the elasticity of labor supply. As the elasticity of labor supply increases, itis more valuable to increase labor income taxation and reduce capital incometaxation, even when we consider the transition process.

The theoretical case against capital income taxation in favor of con-sumption taxation is much stronger than conventionally thought. There arequaliÞers, of course. For example, Hubbard and Judd (1986, 1988) show

7See Gravelle (1994), page 31.


that asset income taxation may be desirable when capital markets are im-perfect. The intuition there is straightforward: capital income taxation maybe useful if it is a substitute for missing capital markets. However, theseÞndings are sensitive to the nature of market incompleteness. It is unclearif those considerations can justify observed capital income tax rates. Forexample, it is difficult to imagine that liquidity constraints could justify thecorporate income tax. It is also possible that capital market failures can bebetter resolved through more modest adjustments of a consumption tax.

We have so far considered the choice between consumption and incometaxation in the simplest possible model: perfect capital markets, perfectcompetition, no risk, and only physical capital and raw labor inputs. Wenow deviate from this simple model and show that the case for consumptiontaxation is strengthened.

2 Imperfect Competition and the BeneÞts of ConsumptionTaxation

Tax reform analyses usually assume perfect competition in all markets. Thisis not a good description of a modern economy. While no one would disagreewith this assertion, it is not immediately clear how this affects tax policyevaluation. We argue that the presence of imperfect competition strengthensthe case for consumption taxation.

The basic idea we pursue here is a combination of two well-known ideas.First, we use the Robinson argument that subsidies can be used to offsetthe distortions if a lump-sum tax is available. At Þrst that seems to be oflimited usefulness since it would imply that most goods would be subsidized,leaving one to wonder what is left to tax to Þnance these subsidies as wellas normal expenditures. Second, Diamond and Mirrlees (1971) tell us thatonly Þnal goods should be taxed, not intermediate goods. Since markupsare similar to taxation, this indicates that the Þnal net tax on intermediategoods should be zero, no matter what the impact on Þnal good taxation.In combination, these principles indicate that Þnal goods should be taxedto Þnance corrective subsidies of any intermediate good, including capitalgoods, which is sold at a price above marginal cost.

We use this controversial assertion in a very limited way. The pure theo-retical argument ignores many practical difficulties. It would be impracticalto construct the perfect corrective policy, and we do not advocate any at-tempt to do so. We make much more limited use of this argument. Weemphasize that if it would be optimal to reduce price-cost margins for in-


termediate goods, then it is surely not a good idea to impose taxes whichaggravate price-cost margins for intermediate goods. In a competitive world,a small tax on intermediate goods may cause only small damage to theeconomy�s efficiency. In a world with imperfect competition in intermedi-ate goods industries, even a small tax on intermediate goods could causesubstantial damage.

Since asset income taxes are equivalent to intermediate good taxes, thisshows that small asset income taxes can create large efficiency losses. Thisobservation strengthens the case for switching away from tax systems, suchas conventional income taxation, which aggravate the distortions of imper-fect competition, and towards consumption tax policies.

2.1 Financing Social Fixed Costs and Taxation

Our results will Þrst appear strange and in conßict with the principles offree market economics. Before we give a more concrete analysis, we presenta simple example of Þxed costs in production where we apply the Diamond-Mirrlees model, and compare its prescriptions to how we actually ÞnanceÞxed costs.

Suppose that there is one capital good, call it computers, which hasconstant unit cost after some large Þxed cost is paid for, say, R&D. Thisgood cannot be produced in a perfectly competitive market since a priceequal to marginal cost will not allow the Þrm to recover the initial Þxedcosts. There must be some deviation from competitive pricing to Þnancethis Þxed cost. Diamond-Mirrlees says that the optimal way to Þnance theÞxed cost for computers is to tax Þnal goods only. The pattern of Þnal goodtaxation is governed by the inverse elasticity rule, not by which goods usecomputers in their manufacture.

Compare this to the manner we actually Þnance Þxed costs, such asR&D expenditures. The computer manufacturer needs to limit competitionso that it can charge a markup over marginal cost sufficient to Þnance theÞxed cost. The economies of scale may be sufficient to deter entry, or perhapsthe computer manufacturer can get a patent on computers. The manner inwhich market power is attained is not particularly important, but some formof market power is necessary.

It has long been recognized that we need to deviate from perfect com-petition if we are to create the proper incentives for innovation. The U.S.Constitution speciÞcally recognizes the need �to promote the progress of sci-ence and useful arts, by securing for limited times to authors and inventorsthe exclusive right to their respective writings and discoveries.� Patents and


copyrights create market power but are valuable instruments to encourageinnovation. Since innovation is considered an important policy concern ofgovernment, it is natural for tax policy be designed so to avoid any interfer-ence with innovation policy.

The analysis in this paper has avoided any explicit modeling of inno-vation. Innovation in a dynamic world has been modeled in many ways;see Judd (1985a) for an example, and Barro and Sala-i-Martin (1995) for areview of the literature. For the sake of simplicity, this paper assumes thatthere is no impact of tax policy on innovation. If we included endogenousinnovation, then moving to a consumption tax would increase innovation incapital goods and further increase our estimates of the gains to consumptiontax reform. The differences would be sensitive to details which are difficultto estimate. In this paper we take a more conservative approach where wefocus solely on the distortions of imperfect competition and the price-costmargins which are easier to estimate.

The incidence of market power in a patent or similar system of protec-tion for intellectual property will likely be very different from the incidenceof the distortions in the ideal Diamond-Mirrlees scheme. Only computerusers pay the markup in computers. Computer users will substitute awayfrom computers and towards alternative intermediate goods. The markupin computer prices will most affect those Þnal goods which use computersin their production. This will result in an inefficient pattern of distortionsacross Þnal goods since those computer-intensive products may not be theones which would be taxed in an optimal Diamond-Mirrlees scheme. It maybe impossible to attain the perfect Diamond-Mirrlees set of distortions, butthis fact only strengthens our case since the excessive burden imposed by im-perfect competition on intermediate goods would only be further aggravatedby any taxation of capital income.

Economic growth requires the creation of some incentives for innovation.The patent and copyright systems succeed in this but they create distortionsin the private sectors. Therefore, tax policy in a modern economy operatesin a world already distorted by other �taxes.� We will see that this insighthas important consequences for the value of consumption taxation.

2.2 Empirical Evidence on Imperfect Competition

We next examine the evidence that there is signiÞcant imperfect competi-tion. There have been many studies of the gap between prices and marginalcosts. Furthermore, the empirical Industrial Organization literature con-tains some industry-speciÞc studies on price-cost margins. These studies


also produce estimates for price-cost margins in the 20% range for somecapital goods (see, e.g., Appelbaum, 1982). Both Hall (1986) and Domowitzet al. (1988) indicate that the margins in the equipment sectors are sub-stantial in size, lying generally between 15% and 40% of the price. Thereis little reason to doubt the presence of signiÞcant economies of scale andsigniÞcant deviations of price over marginal cost. Even a lower estimate of10% is sizable when we remember that it is equivalent to a 10% sales tax onsuch equipment.

Fortunately, our discussion here does not rely critically on these esti-mates of price�cost margins, particularly for investment goods. In par-ticular, R&D expenditures equalled 9.2% of sales for machinery and 4.7%for electrical equipment in 1990. Learning curves also produce increasingreturns to scale which act essentially as a Þxed initial cost. These consider-ations plus a conservative estimate of economies of scale and other long�runÞxed costs puts us in a range relevant for our policy discussions. Therefore,even under conservative readings of the empirical evidence, the importanceof imperfect competition appears substantial.

2.3 A Simple Model of Imperfect Competition

Judd (1995) examines a simple dynamic model which formally establishesour argument. It makes a few key assumptions. First, there is a Þxed numberof goods, all of which are produced in a monopolistically competitive market.Since the number of goods is Þxed, marginal increases in demand results inpure proÞts for all Þrms. Each good can be used for both consumption andinvestment, and each of these goods is used in the production of all goods.Judd (1995) uses a representative agent model with elastic labor supply.

We assume that pure proÞts are taxed at the rate τΠ and that income onmarginal physical investment is taxed at rate τD. One interpretation is thatthe equityholders of each Þrm owns a patent on its good and uses debt toÞnance any physical investment. In equilibrium, the return on equity is thepure rent associated with holding the patent, and debtholders receive themarginal product of the physical investment. Therefore, dividend incomeis subject to corporate level and individual taxation, but the debt-Þnancedphysical capital income is taxed only at the personal level.


2.3.1 The Cost of Capital with Imperfect Competition inCapital Goods Markets

We next illustrate how the social cost of capital is altered by a combina-tion of income taxation and imperfect competition. The cost of capital isdetermined by the usual arbitrage condition. Suppose that a Þrm is con-templating buying one more unit of capital with a social marginal cost ofproduction equal to 1. Because of the markup m by the producer of thecapital good, the investing Þrm pays 1 + m for the unit of capital. Sup-pose that the marginal product of capital is MPK. Assume that the Þrm�sbondholders pay a tax τD on the earnings from this investment and receivean after-tax return of r on alternative investments. Investment will con-tinue until the after-tax return (we assume no depreciation) from a one unitinvestment, MPK(1− τD), equals the opportunity cost of the investment,r(1+m). In equilibrium, the level of investment is determined by

MPK = r1+m

1− τD (4)

If m = 0, (4) is the usual cost of capital formula. In the presence of monop-olistic competition, the upstream markup of m on the purchase of capitalgoods acts in the same way as the downstream taxation of interest income.

To illustrate the combined effects of taxation and imperfect competition,we derive an effective combined tax rate. The situation in (4) is as if therewere no markup and the tax on interest income were equal to τ∗ where

τ∗ = 1− 1− τD1+m

= τD +m

1+m(1− τD) (5)

Table 3 presents values for the total effective tax rate τ∗ for variousvalues of the explicit tax τD and the margin m. For small tax rates andmargins, the total effective tax rate is the sum τD+m. At larger rates, τ

∗ isless than τD +m, but presence of the margin m still substantially increasesthe total distortion. For example, the presence of a 30% margin causes thetotal tax rate to be 38% if τD = .20.

Table 3: Effective Total Tax Rates

τD : .1 .2 .3 .5m :

.05 .14 .24 .33 .52

.10 .18 .27 .36 .54

.20 .25 .33 .42 .58

.30 .31 .38 .46 .62


With the concept of effective total tax in (5), we can see how our earlierarguments apply. First, since markups on capital goods distorts investmentjust as an interest tax would, they produce the same kind of explodingdistortion in (2) which occurs under an interest tax. A uniform markup oncapital goods violates the inverse elasticity principle just as a constant assetincome tax does.

Second, incorporating imperfect competition into our analysis forces usto reconsider the level playing Þeld arguments. The conventional wisdom,based on perfect competition assumptions, is that the 1986 tax changes elim-inated most of the differential taxation of capital goods; Auerbach (1989)is an example of such a study. Even if the explicit income taxes do notdiscriminate among alternative capital goods, the total effective tax rate τ∗

will vary across goods to the extent that their margins vary. Studies suchas Hall and Domowitz et al. both indicate substantial variance in marginsamong capital goods. Since the welfare costs of taxation are increasing in thevariance of inappropriate distortions, our neglect of heterogeneous markupsmake our results conservative estimates of the inefficiency associated withcapital income taxation.

2.3.2 Optimal Tax Policy

We next illustrate what the presence of imperfect competition implies foroptimal tax policy. We assume in this exercise that one can determine themarkups and use them for policy purposes. This is not a realistic assumptionsince it is difficult to measure markups with great precision. The purposeof this exercise is to illustrate how much the presence of markups couldaffect the optimal policy. The results will give us a strong indication of howimportant imperfect competition.

When pure proÞts are taxed at rate τΠ, Judd (1995) shows that thelong-run optimal choice for τD is

τoptD = −m1+ τΠ MEB1+MEB

(6)

where m is the markup of price over marginal cost and MEB is againthe marginal efficiency cost of taxation. If the efficiency cost of taxationis zero then the optimal tax completely neutralizes the monopolistic pricedistortion. This repeats the Robinson Þnding. As in Diamond-Mirrlees, theoptimal tax rate on proÞts, τΠ, is 100%, and the optimal policy eliminatesthe monopolistic price distortion.

While our optimal tax formula (6) is simple, it is not immediately clearthat the desirable subsidy is economically signiÞcant when we use reasonable


values for the markupm, the proÞts tax τΠ, and the marginal excess burdenMEB. We assume m ∈ [.1, .3] as suggested by our discussion of price-costmargins. The range for MEB is taken from Table 2. A key fact is that theequilibrium in our monopolistic competition analysis is essentially the sameas for the competitive model used in Table 2 where τ∗ from (5) is used asthe total effective tax rate on capital income.

Table 4 shows that even if MEB is large, the optimal tax substantiallyreduces the monopolistic distortion. In Table 4 we assume that τΠ = .2, asproposed in the Flat Tax; we arrive at similar conclusions if we use the pureproÞts tax rate implicit in any other major tax reform proposal.

Table 4: Optimal Tax Rates

MEB: .2 .5 1.0 2.0m :

.05 -.04 -.04 -.03 -.02

.10 -.09 -.07 -.06 -.05

.20 -.17 -.15 -.12 -.09

.30 -.26 -.22 -.18 -.14

Table 4 illustrates a number of points. First, the optimal subsidy isnontrivial in most cases. This shows how a system which puts no tax onasset income would still suffer a substantial distortion relative to the ideal.Second, we Þnd that the productive efficiency principle8 is still a good indi-cation of optimal policy even though the proÞts tax is far less than desiredby Diamond-Mirrlees. Third, the desire for productive efficiency is strongeven in cases where the marginal efficiency cost of taxation is high. Theefficiency cost may be high because the revenue need is large or becausethe elasticity of labor supply is high. In either case, tax policy should stillfocus on policies which do not aggravate the pre-existing distortions fromimperfect competition.

The policy implied by Table 4 is impractical. However, the results inTable 4 indicate how far any income tax system is from optimal. Table 4also indicates how concerns about the taxation of pure proÞts are of farless importance than the goal of eliminating productive and intertemporaldistortions.

8Points about productive efficiency would be better demonstrated in a model withheterogeneous capital goods. The more general analysis in Judd (1995) indicates thatthese conclusions are strongly supported in such models.


2.4 BeneÞts of Switching to Consumption Taxation

We next give a quantitative estimate of how monopolistic competition affectsour estimates of the gains from switching to a consumption tax. We continueto use the model in Judd (1995). We Þnd that the estimated beneÞts ofswitching to a ßat consumption tax are substantially increased when weinclude the presence of imperfect competition.

Since price-cost margins are essentially the same as taxes, we can usethe results in Table 2 to draw inferences about the beneÞts of small taxpolicy changes. Suppose that capital goods are sold at 20% above marginalcost. We also will assume that there are labor market imperfections suchas labor unions which cause wage costs to be 10% higher. Then even if theexplicit taxes are τL = τK = .3 initially, the economy really begins withτL = .4 and τK = .5 when we change tax policy. The τL = .4, τK = .5 casein Table 2 then displays the efficiency impact of alternative tax changes ifall marginal proÞts are taxed away. We see that the marginal beneÞts ofreducing asset income taxation are substantially increased, being at leastdoubled and often at least tripled. The magnitudes are uncertain since wedon�t know the values of the critical taste parameters, but the impact ofimperfect competition is clear and substantial for any standard estimate.

Table 2 examined small changes. We next examine large changes in taxpolicy. Table 5 reports the total welfare gain of replacing all income taxationwith consumption taxation. Table 5 measures this gain by expressing thatpercentage change in consumption which is equal to the change in welfarefrom the tax change. For example, in the case of γ = .25, m = 0, andτK = .15, we Þnd that the welfare gain from the switch is equal to animmediate and permanent .12 % increase in consumption9. Table 5 examinescapital income10 tax rates of 15%, 25%, and 35%. The rate τK representsthe marginal tax rate, not the average rate, since the distortion depends onthe marginal tax rate. We examine markups of 0%, 10%, and 20%. Weassume that depreciation is 5% per year and that capital share is 25%.

9This quantity is small, but typical for competitive model. An alternative way toexpress the welfare gain is to report the ratio of welfare gain to the revenue or revenuechange. However, that index is sensitive to the presence of details such as the standarddeduction. The index we use is a cleaner way to express the welfare gains which allows usto ignore details which are not relevant for us.

10We ignore labor taxation since labor is inelastically supplied in our simple analysis.However, the presence of a wage tax with elastic labor supply generally increases thewelfare costs of taxation. Hence, our results are conservative estimates of welfare costs.


Table 5: Welfare Gain (% of Consumption)

τK : .15 .25 .35γ m

.25 .0 .12 .38 .84.1 .37 .79 1.41.2 .54 1.08 1.81

.5 .0 .19 .59 1.30.1 .57 1.21 2.16.2 .81 1.62 2.74

1.1 .0 .24 .76 1.67.1 .72 1.54 2.75.2 1.00 2.04 3.46

Table 5 shows that the presence of a markup substantially increasesthe beneÞts of switching to a consumption tax. In fact, the presence ofjust a 10% markup often doubles the welfare gain relative to the perfectcompetition case. Again we Þnd that these gains are substantial for anyestimate of the critical parameters.

2.5 Asset Pricing Implications

There is substantial interest in the asset pricing implications of tax reform.In particular, a move to a consumption tax would remove the tax burden onnew capital but continue to tax old capital. In a perfectly competitive worldwhere output depends on labor and physical capital alone, competition fromnew capital could lower the market value of old capital. Gravelle (1995)estimates that the Hall-Rabushka Flat Tax would cause a 20-30% fall inthe stock market. This is an important issue which, if true, would createopposition to tax reform.

Gravelle�s estimate assumed perfect competition. However, it is unre-alistic to assume that most equity is associated with perfectly competitiveÞrms. The value of many Þrms consists not only of physical capital but ofintellectual capital. The value of computer software Þrms like Microsoft andpharmaceutical Þrms like PÞzer come from their patents and copyrights, notfrom their physical plant. Competitive entry is made difficult by patents andcopyrights as well as the costs of imitation. While entry may be spurred bythe lower tax rates, the R&D process takes time and will have only delayedeffects on the proÞts of incumbents.

Many Þrms are combinations of physical capital and intellectual capital.Tax reform will reduce the cost of physical capital to each Þrm, causing more


competition among Þrms and lower prices. However, if a Þrm is initiallycharging a price in excess of marginal cost, the increase in demand willincrease his proÞts. Predictions about asset prices need to be changed in animperfectly competitive world. The tax analogy is again apt. Firms withmarket power essentially impose a tax on the product they produce. If thegovernment reduces the tax it imposes on buyers of those products, thenone expects the Þrm to gain pure proÞts through increased demand for itsproduct. For example, if the tax τ in Figure 3 were eliminated, the Þrm�sproÞts would increase by the box Hτm. This gain in proÞts is not presentfor perfectly competitive producers and is ignored in analyses of asset pricechanges which assume perfect competition in the product markets11.

The intuition is clear, and similar to the situation of multiple tax juris-dictions. If the U.S. Federal income tax were repealed and replaced withlump-sum taxation, then output would rise and state income tax revenueswould rise. The same is true here where the producers impose a tax on theirbuyers; a more efficiency U.S. tax system would increase the average Þrm�ssales and increase the revenue from the �tax� it imposes on buyers throughthe gap between price and marginal cost. In the case of a private Þrm, theseextra revenues, current and future, will be immediately capitalized in theÞrm value.

The magnitude of these changes is not so clear. The impact in an openeconomy is clear: if interest rates do not change, then an increase in fu-ture proÞt ßows will immediately increase asset values. However, a radicalchange in U.S. tax policy may produce changes in interest rates in our closedeconomy model. We need to investigate that possibility to establish the ro-bustness of our general claim that the impact of a consumption tax on assetprices is positively affected by the presence of imperfect competition.

We model this explicitly in the model of Judd (1995) with inelastic laborsupply. Essentially, we assume that all goods, Þnal and intermediate, have acommon markup, m. A reduction in the tax on capital income will cause animmediate increase in investment and a gradual increase in aggregate output.There will be ßuctuations in the share of output devoted to consumption andÞnal goods, but that will not affect aggregate asset values since we assumeall goods have the same markup12.

11We always assuming perfect competition in Þnancial markets. We argue that no Þrmembodies a substantial share of all outstanding equity, and no Þrm offers a substantiallyunique risk opportunity.

12In a richer model, different Þrms would sell different goods and experience differentasset price changes. For example, those Þrms who specialize in capital goods would expe-rience an immediate increase in demand whereas those which specialize in consumer goods


Table 6 reports the initial impact on the aggregate market value of equityif we replace an income tax with a lump-sum tax in the model of Judd(1995). In such a tax system, the value of the Þrm would be equal to thereplacement value of its assets if there were no adjustment costs (as weassume) and product markets were perfectly competitive. We assume thatthe economy is initially in the steady state associated with a tax rate τK onall asset income. We also assume that Þxed costs of production are so highthat there are no extranormal proÞts in the initial steady state. We examinethree cases for the intertemporal elasticity of substitution, γ, two values forthe initial income tax rate, τK , and two possible values of the markup, m.

Table 6: Initial Increase in Firm Value(%)

τKγ m .15 .25 .35

.2 .1 1.1 2.0 3.1.2 1.8 3.3 5.2.3 2.3 4.4 6.8

.5 .1 1.8 3.3 5.4- .2 3.0 5.7 9.3

.3 4.1 7.8 12.91.1 .1 2.4 4.5 7.5

.2 4.1 8.0 13.3

.3 5.7 11.3 19.1

Table 6 shows that a transition to a lump-sum tax would increase assetvalues. The impact is small in the cases of small γ because of the slowadjustment in consumption and investment. The case of nearly log utility(γ = 1.1) and a modest markup of 20% implies that the value of the Þrmwould rise in value by 13% if the marginal tax on equity capital were 35%.If a Þrm were Þnanced half by debt and half by debt, all of the increasedvalue would go to equityholders, implying an increase in stock market valueof 26%.

A Flat Tax would be produce different results but its implications areclear. If the Flat Tax rate were 20%, then the value of a perfectly competitiveÞrm would fall by 20% because the expensing provisions create a 20% wedgebetween the value of old and new capital. This is the point Gravelle (1995)makes. However, the change in value of a noncompetitive Þrm would still

would lose sales since the consumption share of output falls in the short-run. However,we assume that all investors are well-diversiÞed, permitting us to focus on aggregate assetvalues.


be increased by the extra pure proÞts it would earn. The net change wouldbe equal to the value in Table 6 minus 20%. Similar arguments apply to aVAT or a national sales tax.

In all cases in Table 6, imperfect competition reduces the negative im-pacts of consumption taxation on the welfare of those, like some elderly,who sell assets to Þnance consumption. Lyon and Merrill (this volume) dis-cusses asset price implications in greater detail. They make points similarto the ones we make here but do not consider equilibrium impacts on salesand interest rates. Their points further reduce any possible asset price re-duction. Our simpler general equilibrium model explores the importance ofimperfect competition and ignores many of the other elements considered byLyons and Merrill. The arguments made here and in Lyons and Merrill rein-force each other, and in combination argue strongly against the pessimisticviews expressed in Gravelle (1995).

2.6 Imperfect Competition and Tax Reform

Imperfect competition is a fact of a modern economy, and should be in-cluded in any tax analysis. In fact, it is the way we provide incentives forinnovation. Capital goods users are already paying a tax in order to Þnancethis investment. While it may not be feasible to relieve that tax burdenthrough tinkering with the tax code, we should still recognize that this pri-vate tax means that further taxation of capital goods substantially damageseconomic efficiency. The presence of this private tax makes it more valuableto move to consumption taxation. The presence of imperfect competitionalso ameliorates the any negative impact on asset prices since the increase inproduction increases pure proÞts in the presence of imperfect competition.Including imperfect competition in our analysis improves the predictions forlong-run growth, beneÞts along the transition, and the immediate impacton asset prices.

3 Risk and Tax Reform

Investment is generally risky, but risk is often ignored in tax reform analy-ses. This section uses Hamilton�s (1986) general equilibrium analysis of thetaxation of risky assets to make some basic points. First, we Þnd that asym-metric treatment of risky assets will affect the equilibrium portfolio of theeconomy. While this is expected, we do this to emphasize the importance ofgeneral equilibrium effects since partial equilibrium analyses lead to contraryconclusions. Second, we emphasize Hamilton�s Þnding that there should be


no differential taxation of risky and safe assets. This example indicates thata goal of tax reform should be to eliminate any distortion between safe andrisky investments. Third, we analyze the utility-revenue trade-off availableto policymakers and demonstrate the importance of incorporating risk inour analysis.

3.1 Asset returns and risk

The most important fact about asset returns in the U.S. is that the annualpre-tax real return to individuals on equity investments has averaged 7%with a standard deviation of 20%, and the mean real return on safe assetshas been 1%. Corporate tax adjustments imply that both mean and varianceshould be 20-40% higher for the risky asset to approximate the opportuni-ties offered to society. The extra return to risky equity is consistent withstandard asset pricing theory, but the magnitude is difficult to explain; seeKocherlakota (1996) for a discussion of asset pricing puzzles. The empiricalpuzzles surrounding asset pricing makes any tax analysis difficult to execute.Even so, we Þnd that including risk in our analysis strengthens the case forconsumption taxation.

3.2 Treatment of Risk in the U.S. Income Tax System

The U.S. tax system appears to discriminate against risky assets and infavor of safe assets. This discrimination depends on the type of investmentand the manner in which an investor holds it. If an asset is held in a deÞnedcontribution pension account, there is no taxation at the personal level.Corporate debt, a relatively safe asset is deducted at the corporate level,implying no taxation of any income generated by such assets. However,income generated by equity investments is taxed at the Þrm level throughthe corporate income tax.

For assets held outside of pension accounts, we need to include personalincome taxation. At the personal level, dividends and interest income aretaxed at the same rate and capital gains have often been taxed at a lowerrate. Since the corporate income tax rate is close to or exceeds the personaltax rate, it appears that risky equity investment held outside of tax-favoredaccounts is taxed at a higher rate than safe debt. These observations indicatethat the current U.S. income tax system produces substantial discriminationagainst risky assets and the investments behind them no matter how theyare held by investors. Hubbard (1993) reviews conventional treatments ofthese issues.


3.3 Hamilton�s Model of Risk and Asset Taxation

There have been many analyses of taxation and risk. Domar and Musgrave(1944) argue that an income tax increases risk taking in the economy. How-ever, the Domar-Musgrave result is substantially altered in a general equi-librium context since many risks faced by the government are passed ontoprivate agents and affect private risk taking. Eaton (1981), Gordon(1985),Hamilton (1986), and Kaplow(1994) have analyzed theoretical issues con-cerning tax systems and risk taking.

Unfortunately, risk is generally absent from quantitative analyses of tax-ation. This is not surprising since it is difficult to incorporate risk in dynamicgeneral equilibrium analysis. It is also unclear how we should calibrate anysuch model since we do not understand why there is such a large gap betweenthe mean return of safe and risky assets. However, we should not totallyignore risk in tax reform analyses. We use Hamilton�s (1986) model to ex-amine the impact of differential taxation since it focuses on the most basicelements of asset allocation and risk. It allows us to compare consumptiontaxation, uniform income taxation, and differential income taxation all inone model.

We assume that there are two types of investment projects. We assumethat the net income from risky assets is taxed at rate τZ and the incomefrom the safe assets is taxed at rate τR. We assume that agents have aconstant relative risk aversion utility function13, and discount the future atrate 4% per year.

We assume that all revenues are rebated lump-sum to investors. Thisis a common assumption we make to abstract from government expendi-ture policies. In this stochastic context, this assumption takes on addedimportance. If revenues were destroyed, then, as Domar and Musgrave haveargued, a constant income tax would shift investment towards the risky as-set. However, we Þnd that assumption to be unrealistic since governmentexpenditures do not immediately react to revenue shocks. The essential ideabehind this assumption is that current revenue shocks lead either to tax cutsin the future, or to increases in government expenditures on goods whichare good substitutes for private consumption. We do not argue that this isthe most valid speciÞcation of actual policies, but use it because it is onewhich allows us to examine the critical issues without modeling Þne detailsof government expenditure policies.

13More precisely, we assume u(c) = c1−1/γ/(1− 1/γ), where γ is also the intertemporalelasticity of consumption used in Table 2.


3.4 Utility and Revenue

We use Hamilton�s model to examine the trade-off between utility and rev-enue. We examine several numerical cases. First, we assume that the riskyasset has a mean return of 10%, a standard deviation of 25%, and thatthe safe asset has a mean return of 1%. We do not defend this particu-lar assumption. In any case, we have recalculated the examples below foralternative means and variances, and Þnd that the qualitative points areunchanged.

Hamilton (1987) examined optimal income taxation in such models. Heshowed that the optimal constant tax policy is to have equal tax rates for safeand risky assets. We examine the global trade-offs among various nonopti-mal tax policies.

Figure 4 displays important features for relative risk aversion of 10 (cor-responding to γ = .1 in Table 2.) This may appear to imply a large riskaversion. However, some implications are reasonable. In particular, the stan-dard deviations of consumption and output are about 1%, which is close toobserved values.

Figure 4 presents two types of curves relating the tax on the safe asset,τR, and the tax on the risky asset, τZ . The curves U.1, U.3, and U.5, areisoutility curves corresponding to the cases where τR = τZ = .1, .3, .5. Thatis, any combination of taxes along U.1 produce the same expected utility asthe tax policy τR = τZ = .1. Expected utility is greater as we move southand west. Similarly, R.1, R.3, and R.5, are the isorevenue curves correspond-ing to the cases where τR = τZ = .1, .3, .5. The dotted line is the 45 degreeline. Revenue increases as we move east and north. A consumption tax isrepresented at the origin where τR = τZ = 0. Note that the isorevenueand isoutility curves are tangent along the 45 degree line, implying that theoptimal policy is one of equal tax rates as predicted by Hamilton.

While the optimality implications of Figure 4 correspond to theory, theglobal trade-offs are strange. Note that revenue is relatively insensitive tochanges in the tax on the risky asset. This is not too surprising since mostwealth is in the safe asset in Figure 4. More surprising is the shape of theisoutility curves away from the optimal policy. We see that if the tax rateon safe assets is much smaller than the tax rate on risky assets, an increasein tax rates can keep utility unchanged or even improve utility!

These features of Figure 4 shows the importance of including uncertaintyexplicitly into our analysis. The normal procedure is to take the averagepre-tax and post-tax returns and insert them into formulas for utility andrevenue in a deterministic model. This approximation will miss incorrectly


Figure 4:

predict the shape of the isoutility curves since it would predict a uniformfall in utility as tax rates rise.

We can use Figure 4 to make some assessments about the value of con-verting to consumption taxation and of other less radical reforms. In Figure4, a constant consumption tax is effectively a lump-sum tax since thereare is no labor supply decision. We proceed under the assumption thatFigure 4 approximates the welfare gain if labor supply were slightly elas-tic. Suppose that τR = .15 and τZ = .35, the situation at point A. Theutility-maximizing policy raising the same revenue is at C, implying a smallreduction in the tax on risky investments and a larger increase in taxationof safe assets. The optimal revenue neutral change is to move to point B,where utility is higher. The move from A to a consumption tax can bedecomposed into two moves, Þrst to a revenue neutral change to a uniformtax at B and then to the origin in Figure 4.

Analyses which ignore the differential taxation of assets will miss theutility gain associated from eliminating nonuniformities, such as the movefrom A to B. This gain would be achieved even if we just integrated corpo-rate and individual taxation. When we add this feature to the analysis, weÞnd another beneÞt from moving to consumption taxation.

Table 7 displays the welfare cost of taxation in the Hamilton model of


risk and taxation. We assume that safe assets are taxed at a rate τR = .1,and that risky assets are taxed at the rates τZ = .1, .4. MEBR is themarginal excess burden of increasing τR, measured as the change in certaintyequivalent consumption per dollar of revenue change. MEBZ (MEBK) isthe marginal excess burden associated with raising a dollar of revenue by asmall increase in the taxation of risky asset income (all asset income). Wealso compute the value of large changes in taxation. To do this, we computethe change in the certainty equivalent of utility, measured in the equivalentconstant consumption ßow. We compare this to the certainty equivalent ofthe change in revenue ßow. The differential burden DBR is the value perdollar of revenue change of eliminating differential taxation. TB is the totalburden of the initial system of taxation per dollar of revenue.

Table 7: Excess Burden of Taxation with Risky Assets

γ τZ τR MEBR MEBZ MEBK DBR TB.5 .1 .1 -.10 -.061 -.056 0 -.025

.4 .1 -1.90 -.510 -.580 -.11 -.220.1 .1 .1 -.01 -.011 -.002 0 -.005

.4 .1 .01 -.120 -.037 -.022 -.029

If τZ = τR = .1 then safe and risky assets are taxed symmetrically. Thetotal burden is small. However, the marginal burden of introducing anyasymmetry is higher than the marginal burden of a uniform increase. In theγ = .5 column, the total burden of taxation is 2.5 cents, and the marginalburden of asset taxation is 5.6 cents. However, the marginal burden ofraising the tax on risky assets only is 6.1 cents. When we examine theasymmetric case of τZ = .4 and τR = .1, the results are more striking.The gain from eliminating all asset income taxation is 22 cents per dollar ofrevenue, but the gain of eliminating just the asymmetric treatment, holdingÞxed total revenue, is 11 cents per dollar of initial revenue. The beneÞts ofreducing asymmetries increases substantially in this case.

These observations even apply to those who hold their equity in 401(k)accounts or similar pension savings accounts. Individual investors still paytaxes on their risky assets through the corporate income tax. In reality, aU.S. taxpayer faces three asset categories - debt, equity, housing - even if hehas all Þnancial assets in tax-favored accounts.

The asymmetric tax treatment of assets produces a substantial burdenon investors in the Hamilton model. We have examined just one particularmodel of taxation and risk, but it is a natural one to study. Further inves-tigation of alternative models would be fruitful, but there is no reason to


suspect that the results would be different. The main intuition is clear: ifthe elasticity of demand for consumption is the same across all states (asis assumed in Hamilton), there is no rationale for asymmetric treatment ofincome across states. The asymmetric treatment of assets by the U.S. taxcode only reduces the efficiency of the U.S. economy.

4 Tax Policy and Human Capital Formation

Human capital is the most important determinant of wealth and income formost individuals and any modern economy. However, income tax analysesdevote less effort to understanding the taxation of human capital than thetaxation of physical capital and labor supply. A separate treatment is neces-sary since human capital is neither just capital nor is it just a feature of thelabor supply. We show that human capital considerations strengthen thecase for consumption taxation, essentially it increases the elasticity of effec-tive labor supply and increases the responsiveness of output to asset income.We also show how it raises new issues about how we should implement aconsumption tax.

4.1 Optimal Taxation of Human Capital and Education In-vestments

Education and other investments in human capital14 present special prob-lems for tax analysis. Education is an investment good since it increaseslabor productivity, but it may also have a consumption value. Diamond-Mirrlees (1971) argues for taxing Þnal goods but against taxation of in-termediate goods. Since human capital appears to be a mixture of laborsupply, investment, and consumption, the implications of these ideas forhuman capital is unclear.

Judd (1999) examines these issues in a dynamic general equilibriummodel. He assumes that individuals invest in both Þnancial assets, A (whichÞnance physical capital, k), and human capital, H. During his life, he earnsrA in asset income where r is the after-tax return on Þnancial assets. Healso earns wL(H,n) in labor income where L(H,n) is effective units of laborinput if he works n hours and his human capital is H, and w is the after-taxwage for a unit of effective labor. He allocates savings between Þnancialinvestments and human capital investment, x. Human capital investments

14There are many forms of investment in human capital. We will focus on education andon-the-job training since they are most relevant for tax analyses. Other forms of humancapital investment, such as child care and medical care, are even more difficult to analyze.


equal x and earn tax credits at the rate s, implying a net cost of x(1 − s).The aggregate production function is f(k,L(H,n)) where f is a standardconstant returns to scale production function.

The incorporation of human capital in this problem generates a tension.If we think of human capital as capital then the logic in Judd (1985b) ar-gues for no taxation of human capital. That leaves labor income as the onlytax in the long-run. However, it is difficult to tax labor income withoutdistorting human capital investments. Judd (1999) shows that if H doesnot affect utility then human capital is purely an intermediate good andthere should be no net taxation on the return of human capital investment,only taxation of hours of labor supply. This can be implemented by taxinglabor income but allowing immediate deduction of all human capital invest-ment expenditures. These results follow exactly the logic of Diamond andMirrlees15.

4.2 Is Education only an Intermediate Good?

IfH is only an intermediate good, then all human capital investments shouldbe expensed. But ifH directly affects utility then the analysis in Judd (1999)shows that we want a positive tax on human capital returns. Many compo-nents of an education appear to have substantial consumption value. Musicappreciation courses in school help one enjoy symphonies and operas laterin life. Sometimes the educational activity itself has both productive valueand aesthetic value. For example, mathematics courses such as calculus,algebra, and topology not only teach the student highly productive skills,but also introduce the student to the beauty of mathematics and the joy ofsolving math problems.

We can get evidence about the character of education by comparingÞnancial returns of alternative assets. If education has a lower Þnancialreturn than comparable Þnancial assets, then human capital must be pro-ducing some nonpecuniary utility returns, is partly a consumption good,and should be taxed. This issue has been addressed somewhat in the liter-ature. Becker (1976) argues that years of education and corporate equityhave roughly the same mean Þnancial return16. Becker (1976) argued (im-

15There have been other analyses of human capital and taxation in economic growthmodels. Jones et al. (1997) argue that there should be no taxation of anything in thelong run. This extreme result arises due to special functional form assumptions made inorder to arrive at a model with a constant growth rate in consumption and all forms ofinvestment. Judd(1999) examines a strictly more general model.

16These are estimates of the social return to education, including any social expenditure


plicitly assuming that education has no Þnal good value) that this showedthat there was no underinvestment in education.

Becker�s comparison with equity raises the question of why education hasas high a risk premium as equity. Previously some economists had arguedthat there was underinvestment since the return on education exceeded thereturn on bonds. Unfortunately, there has been little empirical work on thiswhich considers the risky dimensions of human capital returns. Wage incomemay move with corporate proÞts, but wages are less cyclical than proÞts.Furthermore, the price of risk for human capital depends on the relationshipbetween proÞts and the marginal impact of human capital investment onwage riskiness. Since less educated workers are more likely to experienceunemployment during a recession, education appears to reduce one�s expo-sure to systematic risk. Therefore, the price of risk to be attached to humancapital investments appears to be smaller than that associated with corpo-rate equity. In any case, comparisons with Þnancial assets do not indicateexcessive investment in years of education nor do they indicate any con-sumption component to education. We will proceed under the assumptionthat education is purely an intermediate good.

4.3 The Importance of Human Capital

We next illustrate the quantitative importance of human capital for taxanalyses. We consider a special case of the model described above. Weassume that L(H,n) = Hφn and f(k, L) = kαL1−α. If φ = 0, we havea conventional model with only physical capital. We assume 0 < φ < 1,implying decreasing returns to human capital investments. As in Table 2,let γ > 0 be the elasticity of substitution in consumption and η > 0 theelasticity of labor supply. We highlight the importance of human capital totax analysis by computing the elasticity of long-run output with respect tothe tax and subsidy rates. More precisely, we report percentage change inlong-run output, denoted by εK , εL, and εs in response to a 1% change innet-of-tax rates 1 − τK , and 1 − τL, on physical capital and labor income,and to 1− s, the after-tax cost of human capital investments.

Table 8 reports those elasticities for various values of the critical param-eters. We assume that the level of human capital investment equals half ofphysical capital investment, a conservative choice. We choose φ to be .1 or

as well as the direct monetary and time inputs students. While there has been much effortto reÞne the estimates of the return to years of education, the Becker Þndings are in themiddle range of current estimates, particularly if one adds fringe beneÞts and other non-wage beneÞts of education.


.3. The choice φ = .3 implies that a 10 % increase in accumulated humancapital investments results in a 3% rise in wages, a conservative choice17.Otherwise, we make choose values for γ and η similar to those in Table 2.

Table 8: Elasticity of long-run output to net-of-tax rates

φ = .1 φ = .3γ η.5 .1.5 .5.1 .1

εK εL εs.51 .21 −.11.41 .36 −.09.29 .12 −.07

εL εK εs.75 .78 −.65.55 .84 −.47.36 .37 −.31

Table 8 shows how important it is to include human capital in our analy-sis. We Þrst see that long-run output is sensitive to the treatment of humancapital investments. Even in the φ = .1 case, a 10% change in 1− s changessteady state output by about 1%, and the sensitivity can be half as much asthe sensitivity of output to labor tax changes. For the φ = .3 case, changesin the treatment of human capital investments are as important as changesin the treatment of labor and capital income. The presence of human cap-ital means that the supply of effective labor Hφn is more elastic than ifhuman capital had no marginal value. This increase in labor supply elastic-ity increases the responsiveness of output to asset income taxation and theefficiency cost of capital income taxation.

Table 8 also shows that adding human capital to the analysis increasesthe beneÞts from towards a consumption tax. The sensitivity of output toτK for the φ = .3 is higher than the φ = .1 case, showing that as humancapital becomes more important, the beneÞts from reducing capital incometaxation increases.

To understand the impact of human capital, we examine the allocation oftotal capital across human and physical capital. The equilibrium allocationratio is given by

H

k=φ

α

1

1− τK1− τL1− s (7)

Equation (7) shows that all three tax policy tools affect the H/k ratio inquantitatively symmetric fashions.

Equation (7) also indicates that new issues arise when we convert fromincome to consumption taxation. In an efficient allocation, H/k should just

17These parameter values are also conservative when compared to a common assumptionof φ = 1 in the endogenous growth literature.


equal φ/α, the relative factor share coefficients of human and physical cap-ital. Diamond-Mirrlees says that an optimal tax policy would not deviatefrom this. If we have an income tax where τK = τL but human capital in-vestments are not expensed, s = 0, then the H/k ratio is efficient. However,we saw that the inverse elasticity rule says that τK should be zero. If wejust set τK = 0, the resulting H/k ratio would not be efficient. Efficiencywould be restored if we also set s = τL, which implies that human capitalinvestments should be expensed.

4.4 The Tax Treatment of Human Capital Investments

The U.S. tax code takes a mixed approach to human capital. On-the-jobtraining and a student�s own time are both effectively deductible, while ex-penditures such as tuition and books are generally not deductible. Sinceon-the-job training and students� time comprise most personal direct ex-penditures on human capital investments, some have argued that the U.S.tax system treats human capital well; see Boskin(1977) and Heckman(1976)for discussions of this issue.

However, the picture is more complex. The typical analysis treats thelarge expenditures made by state and local governments on education assubsidies. The Tiebout theory of excludable local publicly provided goodsargues against this view. Local and state education expenditures are Þnancedlargely by local and state taxation and controlled largely by local and statepolitical entities. The Tiebout view argues that the costs of education arecapitalized in the value of land and that public education expenditures areeffectively equivalent to private expenditures. The Tiebout view combinedwith our optimal tax analysis argues that all education expenditures, publicand private, should be deducted from the tax base.

A pure subsidy view of education is also contradicted by the presenceof rationing. Many college students pay tuition far less than the true cost,but only if they meet certain standards. A pure subsidy view ignores thenonprice rationing associated with higher education and the nonprice costswhich are incurred by students competing for those subsidies.

The issue of how to treat educational expenditures is not a minor consid-eration. In fact, 1990 total expenditures on education (other than Federalaid) was $370 billion compared to $576 billion in gross investment in non-residential Þxed capital. Treating educational expenditures as consumptionis similar to taking away all cost recovery from equipment investment, aproposal which would not be regarded as minor.

The Tiebout model is an extreme one, but the main point is robust. In


general, the idea is that citizens of most communities decide to Þnance theeducation of their children together through local taxes. In any rationalmodel of political decisionmaking, these expenditures will respond to theirafter-tax cost. Feldstein and Metcalf (1987) offer evidence that local expen-ditures are affected by Federal Income tax rules. To the extent that stateand local tax deductions affect investment in human capital, some deductionis desirable.

This is currently accomplished partially in the current tax code throughthe deductibility of state and local income and property taxes in the Federalincome tax, but this affects only a part of educational expenditures. Someparents pay substantial non-deductible tuition to send their children to pri-vate schools. Also, itemization is more common among high income families,implying a regressive tax on human capital accumulation. The consumptiontax principle plus the intermediate good view of human capital argues forthe deductibility of all public and private expenditures in all communities.

The Flat Tax (see Hall and Rabushka, 1983), consumption tax (see Brad-ford, 1986), the hybrid tax of McLure and Zodrow(1996), the USA tax (seeWeidenbaum, 1996), and value added (VAT) and national sales tax (NST)proposals all argue for a consumption tax but deÞne �consumption� as in-come minus investment in physical capital only. The various tax proposalsdiffer little on their treatment of human capital investments. The Hall-Rabushka-Armey-Forbes Flat Tax proposals clearly allow few deductionsfor educational investments other than on-the-job training; the sales taxand VAT proposals are similar. The USA tax allows limited deductibility ofsome educational expenses. All would eliminate the deduction for state andlocal taxation which Þnances most educational expenditures. On the otherhand, the Flat Tax would reduce the tax rate on labor income, improvingincentives for human capital investment as indicated in (7).

It is not immediately clear if the current consumption tax proposalshurt or help human capital formation relative to the current tax system.However, it is clear that the treatment of human capital is important. Evenif consumption tax reform does not help human capital directly, Table 8shows that the presence of human capital strengthens the case for reducingthe tax burden on investment in general.

5 Distributional Concerns

We have focused on aggregate output and investment in this paper, andignored distributional concerns. Before ending, we should address theseconcerns. Our main argument is that they are not as severe as they might


appear, and that the human capital issues addressed in the previous sectionsuggest ways to ameliorate some distributional concerns.

5.1 Worker versus Capitalists

A key feature of most current radical tax reforms is the elimination of tax-ation on new investment and reduction of taxation on the current capitalstock, replacing it with taxation on wage income and consumption. This isthe feature which accounts for much of the predicted increase in investmentand output. A shift to wage taxation would seem to hurt workers. Thecounterargument is that the increase in capital accumulation would increaseworker productivity and wages, resulting in higher worker welfare. This ar-gument is often dismissed by consumption tax opponents as too weak, slow,and indirect.

Optimal tax theory presents a very strong argument in favor of elimi-nating investment income taxation. Above we presented the argument whytaxation of asset income in the long-run damages aggregate productivity.This argument also holds when we consider the impact on workers and cap-italists. Judd (1985b) shows that even if the government were in control oftax policy and gave to workers all the receipts from asset income taxation, itwould still choose to have no taxation of asset income in the long-run. Thereis no beneÞt to workers to permanently distort asset accumulation becausethe major effect of long-run asset taxation is to reduce total investment andlabor productivity.

The optimal tax results may appear to be only long run results withlittle force for the foreseeable future. We next investigate the transitionprocess of tax reform by asking how would capitalists and workers share intax reform if we make a small change in the current tax structure. Table 9(taken from Judd, 1984) assumes that the economy18 is in the steady statewith a capital income tax rate of τK and an investment tax credit θ19 andcomputes the change in revenue and wages from a small decrease in τK ora small increase in θ, using increases in wage or consumption taxation toÞnance any shortfall in revenue. Each dR (dW ) entry in Table 9 is thechange, expressed as a percentage of capital income, in the present valueof capital income tax revenue (present value of wage income) caused by

18As in Table 2, we assume a representative agent model with inelastic labor supply.19The investment tax credit proxies for any investment incentive above economic depre-

ciation. In particular, it proxies for accelerated depreciation as well as an explicit ITC.We assume here that the ITC is on all investment, not just equipment. This assumptionis consistent with the nature of consumption tax proposals.


a 1% change in τK or θ. If there were no change in savings, dR wouldequal 1 and dW would equal 0. We examine a variety of values for γ,the intertemporal elasticity of substitution in consumption, and for σ, theelasticity of substitution between capital and labor. We assume that τK isinitially either .3 or .5. We assume that θ is initially .05, representing thepresence of an explicit ITC or accelerated depreciation.

The results in Table 9 address many issues. The values for γ and σ sub-stantially affect the magnitudes of the revenue and income changes. How-ever, we do see some patterns. First, the impact on wages is often substan-tial. It is small only when γ is small, cases where the transition processis slow. In the other cases, a 1 percentage point decrease in τK or a 1percentage point increase in θ increases wages by .4-1.4% of total capitalincome, a substantial change. Also, wages are affected almost equally by a1% change in either τK or θ. Second, changes in θ affect total revenue byless than changes in τK . Therefore, the ITC is a much more potent toolfor increasing wages and labor productivity. This is not surprising since theITC affects only new investment whereas reductions in τK reduce taxationof old capital as well as new investment. Table 9 shows that the ITC canproduce the same improvement for wages at substantially less revenue loss.

Table 9: Disaggregated Effects of Small Tax Changes

τK = .3 τK = .5σ γ

.7 1.00.25

1.0 1.00.25

1.3 1.00.25

Decrease τK Increase θ

dR dW dR dW

-1.02 .93 -.50 1.10-1.01 .60 -.49 .70-.91 .85 -.38 1.00-.95 .50 -.42 .59-.83 .79 -.27 .93-.90 .44 -.37 .52

Decrease τK Increase θ

dR dW dR dW

-1.05 1.35 -.52 1.33-1.03 .87 -.51 .85-.79 1.24 -.28 1.22-.88 .74 -.36 .72-.58 1.16 -.06 1.13-.77 .64 -.25 .62

Third, increases in the ITC could be close to self-Þnancing. The dRnumbers in Table 9 consider only capital income tax revenue. When weadd a reasonable wage tax rate we Þnd that total revenues may rise whenwe increase θ. For example, consider the Þrst line. If τK = .3 initially,then a marginal increase in θ raises before-tax wage income by $1.10 forevery $.50 of revenue loss from capital income taxation. If the marginallabor income tax rate were .45, the extra labor tax revenue would equal$.50 and there would be no net loss in revenue. A labor tax rate of .45is larger than current labor taxation, implying that some increase in labor


taxation would be necessary to balance the budget. If τK = .5 initially,the second set of columns indicate that we only need a .35 marginal taxrate on labor income (a plausible description of the current tax system ifwe include Social Security taxes) for ITC increases to be self-Þnancing. Thepossibility of self-Þnancing ITC increases is not unusual in Table 9. In thecase of Cobb-Douglas technology (σ = 1) and log utility (γ = 1.) we needat most a .38 marginal tax rate on labor income in the low initial tax ratecase. Self-Þnancing decreases in τK are much more unusual, being plausibleonly with a high elasticity of substitution between capital and labor.

In any case, the substantial improvement in before-tax wages means thatwe would need only a small tax rate increase on consumption or wages tobalance the budget. More important, even if they had to pay for an ITCincrease, the workers would be almost always be better off since dR is usuallyless than dW . Only when γ is substantially smaller than the values in Table9 does the revenue loss exceed the wage increases. On the other hand, thisis less likely, albeit not implausible, for reductions in τK .

The analysis in Judd(1984) is biased in favor of consumption taxationbecause of the absence of adjustment costs. However, the estimates in Table9 are conservative since they ignore the elements of imperfect competitionwhich we have argued are important. In particular, if we include imperfectcompetition in our analysis, the τK = .5 case in Table 9 becomes the morerelevant initial condition since Table 3 showed that imperfect competitionsubstantially increases the effective total tax rate on capital income.

Distribution issues are important in making the case for consumptiontaxation. We have seen that the productivity enhancing properties of even asmall movement towards consumption taxation would have beneÞcial effectsfor most even when we consider the transition process.

5.2 Old versus Young

We have used representative agent models of the economy in this paper.This approach ignores intergenerational effects, assuming that all agentslive �forever� and are, effectively, the same age. An alternative paradigmoften used in tax analysis is the overlapping generations (OG) approach.Theoretical analyses, such as Atkinson and Sandmo(1980) have used two-period OGmodels. Auerbach and Kotlikoff(e.g., see Auerbach and Kotlikoff,1987) have used a version where agents live for 55 periods. In such a world,people differ in age, wealth, and planning horizons. Any tax reform couldaffect different cohorts very differently, and affect future generations verydifferently than current generations. The OG approach allows us to analyze


generational issues ignored in representative agent models. We next comparethese approaches and the importance of intergenerational elements for taxpolicy analysis.

The representative agent approach is arguably a good approximation forquestions of aggregate dynamics. It is not that representative agent analy-ses literally assume that agents live forever, nor that agents have perfectlyaltruistic attitudes towards their children. The real question for aggregateanalysis is the relative planning horizon of the typical agent, ßexibility inhis dynamic behavior, and his view of the future. Compare, for example,the classic two-period OG model of Samuelson, and the typical Auerbach-Kotlikoff (AK) model. In the Samuelson model these models, each agentlives for only two periods, youth and old age. If we were to interpret theSamuelson model, we would have to say that each agent at age 20 choosesa constant consumption demand and labor supply for 25 years, and then atage 45 gets to change those levels to other levels which are constant for thenext 25 or so years. Such inßexibility is clearly unrealistic. In the AK ver-sion, each agent is economically active for 55 distinct periods (modeling ages20 through 75), allowed to change consumption and labor decisions in eachyear. The extra ßexibility in the AK version makes it a much more realisticmodel. The extra ßexibility produces much more sensible descriptions of thetransition process after a tax reform and allows us to use empirical analyseswhich similarly assume annual or similarly frequent observations of agents�decisions.

The key difference between the AK model and a representative agentmodel is the length of life of the typical agent, but it is not clear how impor-tant these differences are given the level of discounting typically used. BothAuerbach and Kotlikoff and those who use representative agent models as-sume that agents discount the future at an annual rate of 4% or thereabouts.This implies that a young person at age 20 treats a dollar at age 75 as beingequal to 12 cents at age 20. The difference between the representative agentmodel and the AK model is that the utility derived between ages 20 and 75constitute 88% of lifetime utility for an inÞnitely lived agent and 100% oflifetime utility in the AK model in the absence of a bequest motive.

The representative agent and AK models are similar in their predictionsof aggregate output and dynamics. For any Þxed utility function and pro-duction function, the two models differ, but we know neither tastes nortechnology with precision. The range of predictions of the two models aresimilar once we examine the wide range of empirically sensible taste andtechnology speciÞcations.

The major difference lies in implications for speciÞc individuals. The


advantage of the AK model is that it can be used for intergenerational dis-tribution analyses. That cannot be disputed. Using the AK model, Auer-bach (1996) raises important concerns about the intergenerational impactof tax reform. The problem is that a transition to a consumption tax wouldcause older taxpayers to pay a new tax on their accumulated savings (eitherthrough consumption taxation of the proceeds of asset sales or through afall in the market value of their assets) but that they would not live longenough to enjoy the beneÞts.

However, it is unclear how we should interpret the Auerbach results.Consider, for example, the demographic structure of the AK model. It as-sumes that all die at age 75. These demographic assumptions are inaccurateon two accounts. First, death is an uncertain process20 and many people livelonger than 75 years. Table 10 compares life expectancy in the AK modeland in the U.S. In fact, in the U.S., a 75-year old has a life expectancy of 11years, not 1. When an AK model says that a 75-year-old loses from a taxreform because of a drop in asset prices, that loss presumably arises becausehis life expectancy is just one year. If AK predicts that anyone younger than60 gains, that is presumably because anyone with more than a 15-year lifeexpectancy gains, and that those people gain because any immediate short-run loss is balanced by gains over the following 15 years. When we translatethis interpretation to U.S. demographics, it appears that AK predicts thatanyone younger than 67 would gain from consumption tax reform since a67-year old has about a 16 year life expectancy in the U.S.

20In our discussions we will assume that there is an actuarially fair annuity market.If such markets did not exist then an income tax may be desirable as a way to sharelife expectancy risk. In general, when capital markets are not perfect, income taxationmay dominate consumption taxation. See Hubbard and Judd (1986, 1987, 1988) foranalyses of taxation with capital market imperfections. Future work should integrate theconsiderations of Hubbard and Judd with the concerns of this paper to determine therelative strength of these conßicting forces.


Table 10: Life Expectancy in A-K Model and U.S.

Age A-K Model U.S. Adult Population

Life Fraction Life FractionExpectancy Older Expectancy Older

55 21 .32 25.1 .2960 16 .24 21.1 .2365 11 .16 17.4 .1867 7 .13 16.1 .1670 6 .08 14.1 .1375 1 .01 11.0 .0880 (NA) (NA) 8.3 .04

85+ (NA) (NA) 6.1 .02

Source: Table 119 of the 1998 Statistical Abstract of the U.S.

The second, and more important difference between the AK demographicspeciÞcations and U.S. demographics is the distribution of life expectancy.Suppose that the AK model predicts that all individuals older than 60 loses.That encompasses 24% of the population, a sizable voting block. Auerbach(1996) argues that transition relief to compensate them would substantiallylimit the possible long-run gains from tax reform. This conclusion is notsurprising given the large number of individuals who would be harmed.Also, the AK analysis assumes that a large fraction of the population wouldbe substantially harmed. For example, the 8% of the population in the AKmodel have less than a 6-year life expectancy. It will be particularly difficultto compensate them since the available horizon is so short.

The U.S. demographic situation is not so grim and does not present aslarge a challenge to transition relief. More precisely, only those older than67 have less than a 16-year life expectancy, and they constitute only 16% ofthe population, not the 24% in AK. The fact that the affected population issmaller would make it easier to construct compensatory policies. Also, thereare far fewer who are substantially affected. For example, those with onlya 6-year life expectancy would likely suffer greater losses than the averageloser and constitute 8% of the population in AK but only 2% in the U.S.

This issue becomes even more ambiguous when we incorporate marriageinto the analysis. Suppose a husband and wife have a life expectancy of15 years. Since their deaths are uncertain and somewhat independent, theexpected time until both die would be greater than 15 years. If there issome altruism between husband and wife, then the effective household lifeexpectancy is greater than 15 years.

Our analysis of asset prices with imperfect competition is also relevant


here. We saw that when we add imperfect competition to our analysis,the switch to a consumption tax would reduce asset prices by less than theincrease in pure proÞts to producers. Other features of tax reform reinforcethese points. Auerbach (1996) assumed that there are no unrealized capitalgains. In reality, older taxpayers hold considerable amounts of equity onwhich they have not yet paid capital gains taxes. A consumption tax reformwould forgive the unpaid income taxes on those unrealized capital gains. Infact, when one integrates the reduction in capital gains taxation with anyreasonable fall in asset prices it is quite plausible for an older taxpayer toenjoy a gain in disposable income.

Both the representative agent and overlapping generations models arehighly stylized models with important differences in their demographic struc-ture. Overlapping generations models are capable of analyzing intergenera-tional incidence issues. However, these incidence results will be very sensitiveto the demographic and tax policy details. The conclusions of analyses suchas Auerbach (1996) seem to be overly pessimistic.

5.3 Middle-Income versus Upper-Income

One of the unfortunate features of many consumption tax proposals is thatmiddle-income groups gain relatively little whereas upper-income groupsgain much more in the short-run. The reasons are clear. Middle-incometaxpayers lose key deductions such as the home mortgage interest deduc-tion and the state and local tax deductions. The reduction in asset incometaxation is of less value to them because most of their assets are alreadyin tax-favored vehicles, such as owner-occupied housing and pension fundaccounts. Their ability to shelter asset income is growing under the currentsystem as we increase the scope, size, and liquidity of those special accounts.Upper income groups beneÞt more from the rate reductions and the elimi-nation of asset income taxation since their savings exceed the contributionlimits of pension accounts.

This distribution problem indicates that consumption tax reform need tobe altered in order to form the necessary political coalition. One alternativeis to keep the mortgage interest deduction, but this would be bad newsfor resource allocation. One of the primary beneÞts of the Flat Tax andsimilar proposals is that it would eliminate the current bias towards housinginvestment and against nonresidential business Þxed investment. Since thehousing stock is of roughly the same size as other forms of capital, thisreallocation of investment would substantially improve economic efficiencyin the long run. If we maintain the mortgage interest deduction in the long


run, we would be losing one of the primary beneÞts of the Flat Tax.An alternative idea is to allow some state and local tax deductions,

possibly tied to educational expenditures. This alternative would redirectsome tax relief to the middle class, and is no worse in terms of simplicity thanallowing some form of the mortgage interest deduction. The incidence wouldbe similar to the mortgage interest deduction since both are strongly relatedto income. Allowing some deductions tied to education would be consistentwith the principles of consumption taxation, whereas the mortgage interestdeduction is a clear violation of the conceptual foundations of consumptiontaxation.

Table 11 displays the magnitudes of the estimated tax revenue cost ofvarious deductions in the current tax system. We see that the revenue costof the home mortgage interest deduction is roughly equal to the revenue lossfrom state and local tax deductions for households. The size of these taxexpenditures reßects the current marginal tax rate. The actual revenue losswould be less under a Flat Tax with a marginal rate of 20% or less. Themortgage interest deduction and state and local tax deductions appear tohave roughly the same budgetary consequences. One suspects that they alsohave similar distributional impacts. The beneÞts of the mortgage interestdeduction is perhaps more focused on the middle class since it is capped andsince the top income groups spend less of their income on housing than themiddle class. Of course, one suspects that any mortgage interest deductionincluded in a modiÞed Flat Tax proposal would also be capped, implyingthat a cap on state and local tax deductions would add no greater complexitythan a capped mortgage interest deduction.

Table 11: Major Tax Expenditures, 1998 ($billion, est.)

Home Mortgage Interest Deductions 51.2State and Local Tax Deductions:

Owner-occupied housing 17.7Other nonbusiness deductions 32.1

Source: Table 544, Statistical Abstract of the U.S., 1998

We have focussed on the educational expenditures of state and localgovernment. While education is the major expenditure of state and localgovernment, a deduction tied to those expenditures would be smaller thanthe current state and local tax deduction. We have argued that educationis an intermediate good whether Þnanced privately or through local gov-ernmental entities, and that its tax treatment should not depend on theorganizational form individuals decide to use. This argument suggests that


we ask the same question of other public services, such as police, Þre, andthe judicial system. If they are intermediate goods then they too should beexcluded from a consumption tax base.

A far more detailed examination of the nature and allocation of localpublicly provided goods is needed. Our conjecture is that such an analysiswill produce proposals which deal with transition and distribution problemswithout deviating much from the underlying goals of consumption taxation.

6 Conclusions

Economists have argued that there are large long-run gains from switching toa consumption tax, but some have argued that there are difficult distributionand transition problems. Earlier arguments have been unduly pessimisticbecause they have ignored many important elements. When we includesome of the features of the U.S. economy which makes it a modern andtechnologically advanced economy, such as imperfect competition, humancapital accumulation, and risk, we Þnd that the case for a consumption taxis substantially strengthened.

Imperfect competition is a ubiquitous feature of a modern economy, butit acts as a tax on the U.S. economy. We show that it is particularly dam-aging in the investment goods sector. Innovation in intermediate goods isÞnanced by allowing imperfect competition in intermediate goods indus-tries. This imperfect competition reduces the productive efficiency of theeconomy. Any tax on capital income inßicts even more damage on the incen-tive of the economy to make desirable investments. We show that the gainsfrom eliminating the tax burden on capital income is particularly large.

The current tax system discriminates against risk-taking since equity-Þnanced investments pay more taxes than debt-Þnanced investments. Thisbias has no rational purpose and distorts the allocation of capital. Analyseswhich ignore this feature of the current tax system substantially underesti-mate the value of moving to a consumption tax or more modest integrationproposals.

Human capital is an important part of any modern economy and makeslabor productivity more sensitive to tax policy. Moving to consumption taxwill not only increase investment in physical capital but also increase wagesand the incentive to invest in education and other forms of human capital,producing an even greater increase in long-run output.

These considerations dramatically affect our estimates of the beneÞts ofmoving to a consumption tax. Overall, estimates of the long-run beneÞts


are easily doubled and often tripled by incorporating these elements into ouranalysis.

These new considerations also help with transition problems. Imperfectcompetition effects will push up stock market values, reducing any adverseeffects of tax reform for older taxpayers. We also argue that concerns aboutintergenerational equity are also alleviated when we take a realistic viewof life expectancy. When we consider the role of education as an invest-ment, we see that deductions for educational expenditures may be used toreduce middle-class losses from tax reform without continuing the inefficientpreference for owner-occupied housing.

At a more fundamental level, we argue that a proper understanding of taxsystems show that an income tax is a particularly bad form of consumptiontaxation, and that the current tax system violates most principles of goodtax policy. The choice of tax systems is an important and difficult one, butthe case for good consumption taxation, as embodied in various proposals,is strong and growing stronger.


References

Aaron, Henry J., Harvey Galper, and Joseph Pechman, eds. UneasyCompromise: Problems of a Hybrid Income-Consumption Tax, Brook-ings: Washington, D.C. 1988.

Appelbaum, E. �The Estimation of the Degree of Oligopoly Power,�Journal of Econometrics, 19 (1982): 287�299.

Atkinson, Anthony B., and Agnar Sandmo. �Welfare Implicationsof the Taxation of Savings,� The Economic Journal 90 (September,1980).

Atkinson, A., and Stiglitz, J. �The Structure of Indirect Taxation andEconomic Efficiency.� Journal of Public Economics 1 (1972), 97�119.

Auerbach, A. J. �The Optimal Taxation of Heterogeneous Capital,�Quarterly Journal of Economics 93 (1979), 589-612.

Auerbach, A.J., and L. Kotlikoff. Dynamic Fiscal Policy, Cambridge:Cambridge University Press, 1987

Auerbach, Alan J. �The Deadweight Loss from �Non-neutral� CapitalIncome Taxation,� Journal of Public Economics 40 (October, 1989):1-36.

Auerbach, Alan J. �Tax Reform, Capital Accumulation, Efficiency,and Growth,� in: Henry J. Aaron and William G. Gale, eds., Eco-nomic Effects of Fundamental Tax Reform, The Brookings Institution:Washington, D.C., 1996.

Balcer, Y., I. GarÞnkel, K. Krynski, and E. Sadka. in Social PolicyEvaluation: An Economic Perspective, E. Helpman, A. Razin, and E.Sadka, eds. Academic Press: New York, 1983.

Barro, Robert J., and Xavier Sala-i-Martin. Economic Growth,McGraw-Hill: New York, 1995.

Becker, Gary. Human Capital: A Theoretical and Empirical Analysis.Chicago and London: University of Chicago Press, 1976.

Boskin, M. �Notes on the Tax Treatment of Human Capital.� In U.S.Department of Treasury Office of Tax Analysis, Conference on TaxResearch 1975. Washington, D. C: U.S. Government Printing Office,1977, pages 185-195.


Bradford, D. Untangling the Income Tax , Cambridge: Harvard Uni-versity Press, 1986

Diamond, P. A. �Taxation and Public Production In a Growth Set-ting.� In Models of Economic Growth, J. A. Mirrlees and N. H. Stern(eds.), 215-234, London, Macmillan, 1973.

Diamond, P. A. and Mirrlees, J. A. �Optimal Taxation and PublicProduction,� American Economic Review, 61 (1971), 8-27 and 261-278.

Domar, Evsey D., and Richard A. Musgrave. �Proportional IncomeTaxation and Risk-Taking.� Quarterly Journal of Economics 58 (May,1944): 388�422.

Domowitz, Ian, R. Glenn Hubbard, and Bruce C. Petersen, �BusinessCycles and the Relationship between Concentration and Price-CostMargins,� Rand Journal of Economics 17 (Spring 1986a), 1�17.

Domowitz, Ian, R. Glenn Hubbard, and Bruce C. Petersen, �TheIntertemporal Stability of the Concentration-Margins Relationship,�Journal of Industrial Economics 35 (Sept. 1986b), 13�34.

Eaton, J. �Fiscal Policy, Inßation and the Accumulation of Risky Cap-ital,� Review of Economic Studies 48 (1981), 435-45.

Hall, R. E. and Rabushka, A. Low Tax, Simple Tax, Flat Tax,(McGraw-Hill, New York), 1983

Feldstein, M. S. �The Rate of Return, Taxation and Personal Savings,�Economic Journal 88 (1978) 482-487.

Feldstein, M. S., and Metcalf, G. E. �The Effect of Federal Tax De-ductibility on State and Local Taxes and Spending,� Journal of Polit-ical Economy 95 (1987), 710�736

Gordon, R. H. �Taxation of Corporate Capital Income: Tax Revenuesversus Tax Distortions,� Quarterly Journal of Economics 100 (1985):1-27.

Goulder, L. H., and P. Thalmann. �Approaches to Efficient CapitalTaxation: Leveling the Playing Field vs. Living by the Golden Rule,�Journal of Public Economics 50 (February, 1993): 169-96.


Gravelle, J. The Economic Effects of Taxing Capital Income, Cam-bridge: MIT Press. 1994.

Gravelle, Jane. �The Flat Tax and Other Tax Proposals: Who WillBear the Tax Burden?� CRS Report for Congress 95-1141E, November29, 1995.

Hall, Robert E. and Alvin Rabushka. Low Tax, Simple Tax, Flat Tax ,(McGraw-Hill, New York), 1983.

Hall, Robert E. �Market Structure and Macro Fluctuations,� Brook-ings Papers on Economic Activity, (2, 1986), 285-322

Hamilton, J. H., 1987. �Taxation, Savings, and Portfolio Choice in aContinuous Time Model.� Public Finance 42, 264-82.

Heckman, James. �A Life-Cycle Model of Earnings, Learning , andConsumption,� Journal of Political Economy, , 84(4, August 1976),pp. S11-S44.

Hubbard, R. Glenn, and Kenneth L. Judd. �Liquidity Constraints, Fis-cal Policy, and Consumption� Brookings Papers on Economic Activity,v0 (1, 1986), pp. 1-50.

Hubbard, R. Glenn, and Kenneth L. Judd. �Social Security and Indi-vidual Welfare: Precautionary Saving, Borrowing Constraints, and thePayroll Tax,� American Economic Review 77 (4, September, 1987),pp. 630-46.

Hubbard, R. Glenn, and Kenneth L. Judd. �Capital Market Imper-fections and Tax Policy Analysis in the Life Cycle Model,� Annalesd�Economie et de Statistique v0(9, Jan.-Mar, 1988), pp. 111-39.

Jones, L., R. Manuelli, and P. Rossi. �Optimal Taxation in Models ofEndogenous Growt,� Journal of Political Economy 101(3, June 1993),pp. 485-517.

Jorgenson, Dale W, and Kun Young Yun. �Tax Reform and U.S. Eco-nomic Growth,� Journal of Political Economy; 98 (Number 5, Octo-ber, 1990), S151-93.

Judd, Kenneth L. �On the Performance of Patents,� Econometrica 53(3, May 1985), pp. 567-85.


Judd, Kenneth L. �Redistributive and Efficiency Effects of Capital TaxChanges in a Dynamic Model,� mimeo, 1984, Northwestern University.

Judd, Kenneth L. �Redistributive Taxation in a Simple Perfect Fore-sight Model,� Journal of Public Economics 28 (October, 1985):59-83.

Judd, Kenneth L. �The Welfare Cost of Factor Taxation in a PerfectForesight Model,� Journal of Political Economy 95 (August, 1987).

Judd, Kenneth L. �The Optimal Tax on Capital Income is Negative,�Hoover Institution, mimeo, 1995.

Judd, Kenneth L. �Optimal Taxation and Spending in General Com-petitive Growth Models,� Journal of Public Economics (1999).

Kaplow, Louis. �Taxation and risk taking: a general equilibrium per-spective.� National Tax Journal 47 (1994): 789-798.

Kocherlakota, Narayana R. �The Equity Premium: It�s Still a Puzzle,�Journal of Economic Literature, 34 (March, 1996), 42-71.

McLure, C., and Zodrow, G. R., 1996. �A Hybrid Approach to theDirect Taxation of Consumption.� In Frontiers of Tax Reform, editedby Michael Boskin, Hoover Press: Stanford, CA.

Robinson, Joan. The Economics of Imperfect Competition, London:MacMillan. 1934.

Stiglitz, J. E., and Dasgupta, P., 1971. �Differential Taxation, PublicGoods, and Economic Efficiency,� Review of Economic Studies 38,151-174.

U. S. Department of the Treasury, Integration of the Individual andCorporate Tax Systems: Taxing Business Income Once. Washington,D. C.: U. S. Government Printing Office, 1992.

Weidenbaum, M., 1997. �The Nunn-Domenici USA Tax,� In Frontiersof Tax Reform, edited by Michael Boskin, Hoover Press: Stanford, CA

Statistical Abstract of the United States: 1998. U.S. Department ofCommerce.

the impact of tax reform in modern dynamic …judd/papers/aeifin.pdfthe impact of tax reform in...

Documents